diff --git a/.github/workflows/format.yml b/.github/workflows/format.yml
index bd179cb..fe02d66 100644
--- a/.github/workflows/format.yml
+++ b/.github/workflows/format.yml
@@ -21,13 +21,8 @@ jobs:
   actions:
     runs-on: ubuntu-latest
     steps:
-      # Dogfood this repo's own action: PR events run the PR's action.yml, issue events run main's
-      - uses: actions/checkout@v7
-        with:
-          ref: ${{ github.event.pull_request.head.sha || github.sha }}
-
       - name: Run Ultralytics Actions
-        uses: ./
+        uses: ultralytics/actions@main
         with:
           token: ${{ secrets._GITHUB_TOKEN || github.token }} # Auto-generated fallback
           labels: true # Auto-label issues/PRs using AI
diff --git a/AGENTS.md b/AGENTS.md
index 3eb92b8..f51c628 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -66,7 +66,7 @@ Key flow: GitHub workflow event → `action.yml` step (gated by `github.event_na
 
 Most shared utilities are re-exported through `actions/utils/__init__.py` — keep `__all__` updated when adding exports.
 
-Self-hosting detail: `.github/workflows/format.yml` checks out the event's ref (PR head for pull requests, main for issue events) and runs `uses: ./` — so PRs here dogfood both the Python package and `action.yml` itself before merge.
+Security detail: `.github/workflows/format.yml` runs `ultralytics/actions@main` because it receives write credentials and AI secrets; never execute a PR checkout as a local action in that workflow. PR package changes are exercised by the test workflow before merge.
 
 ## Conventions
 
diff --git a/README.md b/README.md
index 03088d6..b141ec2 100644
--- a/README.md
+++ b/README.md
@@ -186,7 +186,7 @@ Check every pull request commit author against the central Ultralytics CLA signa
 
 ### 6. Setup uv Action
 
-Install the latest uv with retry support and optionally activate a Python environment.
+Use the official uv setup action with Ultralytics defaults, optional Python environment activation, and dependency caching.
 
 ```yaml
 - uses: ultralytics/actions/setup-uv@main
diff --git a/README.zh-CN.md b/README.zh-CN.md
index 7ef5dab..143b3d6 100644
--- a/README.zh-CN.md
+++ b/README.zh-CN.md
@@ -186,7 +186,7 @@ jobs:
 
 ### 6. Setup uv Action
 
-安装最新版 uv 并支持重试，同时可选择激活 Python 环境。
+使用带 Ultralytics 默认配置的官方 uv setup action，并可选择激活 Python 环境和启用依赖缓存。
 
 ```yaml
 - uses: ultralytics/actions/setup-uv@main
diff --git a/action.yml b/action.yml
index 85c3ab1..f75c45c 100644
--- a/action.yml
+++ b/action.yml
@@ -11,7 +11,7 @@ inputs:
     description: "GitHub token"
     required: true
   labels:
-    description: "Run issue and PR auto-labeling"
+    description: "Run issue and PR auto-labeling with first responses"
     required: false
     default: "false"
   python:
@@ -112,45 +112,38 @@ runs:
       # Note tomli required for codespell with pyproject.toml
       env:
         INPUTS_SPELLING: ${{ inputs.spelling }}
-        GITHUB_REPOSITORY: ${{ github.repository }}
-        GITHUB_SHA: ${{ github.sha }}
+        ACTION_PATH: ${{ github.action_path }}
       run: |
         echo "::group::Install Dependencies"
         trap 'echo "::endgroup::"' EXIT
-        # Source tarballs install ~2x faster than git+ URLs, which clone the repository first. Test the current commit
-        # in the ultralytics/actions repo, otherwise install main.
-        if [ "$GITHUB_REPOSITORY" = "ultralytics/actions" ]; then
-          echo "Installing from commit: $GITHUB_SHA"
-          packages="https://github.com/ultralytics/actions/archive/${GITHUB_SHA}.tar.gz"
-        else
-          packages="https://github.com/ultralytics/actions/archive/refs/heads/main.tar.gz"
-        fi
-
+        package_dir=$(mktemp -d "$RUNNER_TEMP/ultralytics-actions.XXXXXX")
+        uv build --wheel "$ACTION_PATH" --out-dir "$package_dir"
+        packages=("$package_dir"/*.whl)
         if [ "$INPUTS_SPELLING" = "true" ]; then
-          packages="$packages codespell tomli"
+          packages+=(codespell tomli)
         fi
         # On macOS, don't use sudo as it can cause environment issues
         if [ "$(uname)" = "Darwin" ]; then
-          uv pip install --system --break-system-packages $packages
+          uv pip install --system --break-system-packages "${packages[@]}"
         else
-          sudo env "PATH=$PATH" uv pip install --system --break-system-packages $packages
+          sudo env "PATH=$PATH" uv pip install --system --break-system-packages "${packages[@]}"
         fi
         ultralytics-actions-info
       shell: bash
 
     # Checkout Repository ----------------------------------------------------------------------------------------------
     - name: Checkout Repository
-      if: github.event.action == 'opened' || github.event.action == 'synchronize' || github.event.action == 'review_requested'
+      if: github.event_name == 'pull_request' && (github.event.action == 'opened' || github.event.action == 'synchronize' || github.event.action == 'review_requested')
       uses: actions/checkout@v7
       with:
-        repository: ${{ github.event.pull_request.head.repo.full_name || github.repository }}
+        repository: ${{ github.event.pull_request.head.repo.full_name }}
         token: ${{ inputs.token }}
-        ref: ${{ github.head_ref || github.ref }}
+        ref: ${{ github.event.pull_request.head.ref }}
         fetch-depth: 2
 
     # File header ------------------------------------------------------------------------------------------------------
     - name: File Header
-      if: (github.event_name == 'pull_request' || github.event_name == 'pull_request_target') && (github.event.action == 'opened' || github.event.action == 'synchronize')
+      if: github.event_name == 'pull_request' && (github.event.action == 'opened' || github.event.action == 'synchronize')
       env:
         HEADER: ${{ inputs.header }}
       run: |
@@ -174,7 +167,7 @@ runs:
     # D413: Missing blank line after last section
     # --target-version is Python 3.8 for --extend-select UP (pyupgrade)
     - name: Run Python
-      if: (github.event_name == 'pull_request' || github.event_name == 'pull_request_target') && inputs.python == 'true' && (github.event.action == 'opened' || github.event.action == 'synchronize')
+      if: github.event_name == 'pull_request' && inputs.python == 'true' && (github.event.action == 'opened' || github.event.action == 'synchronize')
       run: |
         echo "::group::Run Python"
         trap 'echo "::endgroup::"' EXIT
@@ -196,7 +189,7 @@ runs:
 
     # Biome (JavaScript, TypeScript, JSON, JSX, TSX) ------------------------------------------------------------------
     - name: Run Biome
-      if: (github.event_name == 'pull_request' || github.event_name == 'pull_request_target') && inputs.biome == 'true' && (github.event.action == 'opened' || github.event.action == 'synchronize')
+      if: github.event_name == 'pull_request' && inputs.biome == 'true' && (github.event.action == 'opened' || github.event.action == 'synchronize')
       run: |
         echo "::group::Run Biome"
         trap 'echo "::endgroup::"' EXIT
@@ -212,7 +205,7 @@ runs:
 
     # Prettier (JavaScript, JSX, Angular, Vue, Flow, TypeScript, CSS, HTML, JSON, GraphQL, Markdown, YAML) -------------
     - name: Run Prettier
-      if: (github.event_name == 'pull_request' || github.event_name == 'pull_request_target') && (inputs.prettier == 'true' || inputs.markdown == 'true') && (github.event.action == 'opened' || github.event.action == 'synchronize')
+      if: github.event_name == 'pull_request' && (inputs.prettier == 'true' || inputs.markdown == 'true') && (github.event.action == 'opened' || github.event.action == 'synchronize')
       run: |
         echo "::group::Run Prettier"
         trap 'echo "::endgroup::"' EXIT
@@ -233,7 +226,7 @@ runs:
 
     # Swift formatting -------------------------------------------------------------------------------------------------
     - name: Run Swift Formatter
-      if: (github.event_name == 'pull_request' || github.event_name == 'pull_request_target') && inputs.swift == 'true' && (github.event.action == 'opened' || github.event.action == 'synchronize')
+      if: github.event_name == 'pull_request' && inputs.swift == 'true' && (github.event.action == 'opened' || github.event.action == 'synchronize')
       run: |
         echo "::group::Run Swift Formatter"
         trap 'echo "::endgroup::"' EXIT
@@ -244,13 +237,13 @@ runs:
 
     # Dart formatting --------------------------------------------------------------------------------------------------
     - name: Setup Dart SDK
-      if: (github.event_name == 'pull_request' || github.event_name == 'pull_request_target') && inputs.dart == 'true' && (github.event.action == 'opened' || github.event.action == 'synchronize')
+      if: github.event_name == 'pull_request' && inputs.dart == 'true' && (github.event.action == 'opened' || github.event.action == 'synchronize')
       uses: dart-lang/setup-dart@v1
       with:
         sdk: stable
 
     - name: Run Dart Formatter
-      if: (github.event_name == 'pull_request' || github.event_name == 'pull_request_target') && inputs.dart == 'true' && (github.event.action == 'opened' || github.event.action == 'synchronize')
+      if: github.event_name == 'pull_request' && inputs.dart == 'true' && (github.event.action == 'opened' || github.event.action == 'synchronize')
       run: |
         echo "::group::Run Dart Formatter"
         trap 'echo "::endgroup::"' EXIT
@@ -261,7 +254,7 @@ runs:
     # Spelling ---------------------------------------------------------------------------------------------------------
     # --ignore-regex '\b[a-z]+[A-Z][a-zA-Z]*\b' ignores camelCase words (e.g. repoRes, dateA) to prevent false positives
     - name: Run Codespell
-      if: (github.event_name == 'pull_request' || github.event_name == 'pull_request_target') && inputs.spelling == 'true' && (github.event.action == 'opened' || github.event.action == 'synchronize')
+      if: github.event_name == 'pull_request' && inputs.spelling == 'true' && (github.event.action == 'opened' || github.event.action == 'synchronize')
       run: |
         echo "::group::Run Codespell"
         trap 'echo "::endgroup::"' EXIT
@@ -291,9 +284,12 @@ runs:
       shell: bash
       continue-on-error: true
 
-    # Autolabel Issues and PRs (labels and first interaction) ---------------------------------------------------------
-    - name: Autolabel Issues and PRs
-      if: inputs.labels == 'true' && (github.event.action == 'opened' || github.event.action == 'created')
+    # PR automation and first interaction ------------------------------------------------------------------------------
+    - name: PR Automation and First Interaction
+      if: |
+        (inputs.labels == 'true' && (github.event.action == 'opened' || github.event.action == 'created')) ||
+        ((github.event_name == 'pull_request' || github.event_name == 'pull_request_target') &&
+         github.event.action == 'opened' && (inputs.summary == 'true' || inputs.review == 'true'))
       env:
         GITHUB_TOKEN: ${{ inputs.token }}
         FIRST_ISSUE_RESPONSE: ${{ inputs.first_issue_response }}
@@ -302,9 +298,11 @@ runs:
         ANTHROPIC_API_KEY: ${{ inputs.anthropic_api_key }}
         MODEL: ${{ inputs.model }}
         REVIEW_MODEL: ${{ inputs.review_model }}
+        LABELS: ${{ inputs.labels }}
+        SUMMARY: ${{ inputs.summary }}
         REVIEW: ${{ inputs.review }}
       run: |
-        echo "::group::Autolabel Issues and PRs"
+        echo "::group::PR Automation and First Interaction"
         trap 'echo "::endgroup::"' EXIT
         ultralytics-actions-first-interaction
       shell: bash
@@ -331,7 +329,7 @@ runs:
 
     # Bun Update ---------------------------------------------------------------------------------------------------------
     - name: Bun Update
-      if: (github.event_name == 'pull_request' || github.event_name == 'pull_request_target') && inputs.bun_update == 'true' && github.event.action == 'opened'
+      if: github.event_name == 'pull_request' && inputs.bun_update == 'true' && github.event.action == 'opened'
       run: |
         echo "::group::Bun Update"
         trap 'echo "::endgroup::"' EXIT
@@ -359,25 +357,25 @@ runs:
       # Auto-format commit on the opened event. The recursion risk starts after that commit pushes a synchronize
       # event as github_username, so keep the bot guard only for synchronize events.
       if: |
-        (github.event_name == 'pull_request' || github.event_name == 'pull_request_target') &&
+        github.event_name == 'pull_request' &&
         (github.event.action == 'opened' || github.event.action == 'synchronize') &&
-        (github.event.action == 'opened' || github.actor != inputs.github_username)
+        (github.event.action == 'opened' || github.actor != inputs.github_username) &&
+        github.event.pull_request.head.repo.full_name == github.repository
       env:
         GITHUB_USERNAME: ${{ inputs.github_username }}
         GITHUB_EMAIL: ${{ inputs.github_email }}
         PR_HEAD_REF: ${{ github.event.pull_request.head.ref }}
-        PR_HEAD_SHA: ${{ github.event.pull_request.head.sha }}
       run: |
         echo "::group::Commit and Push Changes"
         trap 'echo "::endgroup::"' EXIT
         COMMIT_MSG="Auto-format by https://ultralytics.com/actions"
-        git config --global user.name "$GITHUB_USERNAME"
-        git config --global user.email "$GITHUB_EMAIL"
+        git config user.name "$GITHUB_USERNAME"
+        git config user.email "$GITHUB_EMAIL"
         git add .
         git reset HEAD -- .github/workflows/  # workflow changes are not permitted with default token
         if ! git diff --staged --quiet; then
           git commit -m "$COMMIT_MSG"
-          git push --force-with-lease="refs/heads/${PR_HEAD_REF}:${PR_HEAD_SHA}" origin "HEAD:refs/heads/${PR_HEAD_REF}"
+          git push origin "HEAD:refs/heads/${PR_HEAD_REF}"
         else
           echo "No changes to commit"
         fi
@@ -386,7 +384,7 @@ runs:
 
     # Broken links -----------------------------------------------------------------------------------------------------
     - name: Broken Link Checker
-      if: inputs.links == 'true' && github.event.action != 'closed'
+      if: inputs.links == 'true' && github.event_name == 'pull_request' && github.event.action != 'closed'
       uses: lycheeverse/lychee-action@v2.9.0
       with:
         # Check all markdown and html files in repo. Ignores the following status codes to reduce false positives:
diff --git a/actions/__init__.py b/actions/__init__.py
index 01a87ba..e9e67eb 100644
--- a/actions/__init__.py
+++ b/actions/__init__.py
@@ -28,4 +28,4 @@
 #     ├── test_summarize_pr.py
 #     └── ...
 
-__version__ = "0.3.1"
+__version__ = "0.3.2"
diff --git a/actions/first_interaction.py b/actions/first_interaction.py
index 337bc71..a67f47f 100644
--- a/actions/first_interaction.py
+++ b/actions/first_interaction.py
@@ -3,12 +3,12 @@
 from __future__ import annotations
 
 import os
-import time
 
 from . import review_pr
 from .summarize_pr import SUMMARY_MARKER
 from .utils import (
     ACTIONS_CREDIT,
+    GITHUB_API_URL,
     Action,
     filter_labels,
     format_skipped_files_dropdown,
@@ -16,8 +16,11 @@ from .utils import (
     get_response,
     remove_html_comments,
 )
+from .utils.common_utils import check_links_in_string
 
 BLOCK_USER = os.getenv("BLOCK_USER", "false").lower() == "true"
+AUTO_LABELS = os.getenv("LABELS", "true").lower() == "true"
+AUTO_PR_SUMMARY = os.getenv("SUMMARY", "true").lower() == "true"
 AUTO_PR_REVIEW = os.getenv("REVIEW", "true").lower() == "true"
 
 
@@ -31,7 +34,7 @@ def apply_and_check_labels(event, number, node_id, issue_type, username, labels,
     if normalized := [available.get(label.lower(), label) for label in labels if label.lower() in available]:
         print(f"Applying labels: {normalized}")
         event.apply_labels(number, node_id, normalized, issue_type)
-        if "Alert" in normalized and not event.is_org_member(username):
+        if any(label.lower() == "alert" for label in normalized) and not event.is_org_member(username):
             event.handle_alert(number, node_id, issue_type, username, block=BLOCK_USER)
 
 
@@ -44,8 +47,7 @@ def get_event_content(event) -> tuple[int, str, str, str, str, str, str]:
         item = data["issue"]
         issue_type = "issue"
     elif name in ["pull_request", "pull_request_target"]:
-        pr_number = data["pull_request"]["number"]
-        item = event.get_repo_data(f"pulls/{pr_number}")
+        item = data["pull_request"]
         issue_type = "pull request"
     elif name == "discussion":
         item = data["discussion"]
@@ -61,102 +63,54 @@ def get_event_content(event) -> tuple[int, str, str, str, str, str, str]:
     return number, node_id, title, body, username, issue_type, action
 
 
-def get_relevant_labels(
-    issue_type: str, title: str, body: str, available_labels: dict, current_labels: list
-) -> list[str]:
-    """Determines relevant labels for GitHub issues/discussions using AI."""
-    filtered_labels = filter_labels(available_labels, current_labels, is_pr=(issue_type == "pull request"))
+def get_first_interaction_response(
+    event,
+    issue_type: str,
+    title: str,
+    body: str,
+    username: str,
+    available_labels: dict,
+    current_labels: list,
+    repository_context: str,
+) -> dict:
+    """Generate labels and a first response for an issue or discussion in one LLM call."""
+    issue_discussion_response = f"""
+👋 Thanks @{username} for opening this `{event.repository}` {issue_type}. This is an automated first response; an Ultralytics engineer will assist soon.
+"""
+
+    configured_example = os.getenv("FIRST_ISSUE_RESPONSE")
+    example = configured_example or issue_discussion_response
+    org_name, repo_name = event.repository.split("/")
+    filtered_labels = filter_labels(available_labels, current_labels)
     labels_str = "\n".join(f"- {name}: {description}" for name, description in filtered_labels.items())
 
-    prompt = f"""Select the top 1-3 most relevant labels for the following GitHub {issue_type}.
-
-INSTRUCTIONS:
-1. Review the {issue_type} title and description.
-2. Consider the available labels and their descriptions.
-3. Choose 1-3 labels that best match the {issue_type} content.
-4. Only use the "Alert" label when you have high confidence that this is an inappropriate {issue_type}.
-5. Respond ONLY with the chosen label names (no descriptions), separated by commas.
-6. If no labels are relevant, respond with 'None'.
-{'7. Only use the "bug" label if the user provides a clear description of the bug, their environment with relevant package versions and a minimum reproducible example.' if issue_type == "issue" else ""}
-
-AVAILABLE LABELS:
-{labels_str}
-
-{issue_type.upper()} TITLE:
-{title}
-
-{issue_type.upper()} DESCRIPTION:
-{body[:16000]}
-
-YOUR RESPONSE (label names only):
-"""
-    messages = [
-        {
-            "role": "system",
-            "content": "You are an Ultralytics AI assistant that labels GitHub issues, PRs, and discussions.",
-        },
-        {"role": "user", "content": prompt},
-    ]
-    suggested_labels = get_response(messages, temperature=1.0)
-    if "none" in suggested_labels.lower():
-        return []
-
-    available_labels_lower = {name.lower(): name for name in filtered_labels}
-    return [
-        available_labels_lower[label.lower().strip()]
-        for label in suggested_labels.split(",")
-        if label.lower().strip() in available_labels_lower
-    ]
-
-
-def get_first_interaction_response(event, issue_type: str, title: str, body: str, username: str) -> str:
-    """Generates a custom LLM response for GitHub issues or discussions (NOT PRs - PRs use unified call)."""
-    issue_discussion_response = f"""
-👋 Hello @{username}, thank you for submitting a `{event.repository}` 🚀 {issue_type.capitalize()}. To help us address your concern efficiently, please ensure you've provided the following information:
-
-1. For bug reports:
-   - A clear and concise description of the bug
-   - A minimum reproducible example [MRE](https://docs.ultralytics.com/help/minimum-reproducible-example/) that demonstrates the issue
-   - Your environment details (OS, Python version, package versions)
-   - Expected behavior vs. actual behavior
-   - Any error messages or logs related to the issue
-
-2. For feature requests:
-   - A clear and concise description of the proposed feature
-   - The problem this feature would solve
-   - Any alternative solutions you've considered
-
-3. For questions:
-   - Provide as much context as possible about your question
-   - Include any research you've already done on the topic
-   - Specify which parts of the [documentation](https://docs.ultralytics.com/), if any, you've already consulted
-
-Please make sure you've searched existing {issue_type}s to avoid duplicates. If you need to add any additional information, please comment on this {issue_type}.
-
-Thank you for your contribution to improving our project!
-"""
-
-    example = os.getenv("FIRST_ISSUE_RESPONSE") or issue_discussion_response
-    org_name, repo_name = event.repository.split("/")
-
-    prompt = f"""Generate a customized response to the new GitHub {issue_type} below:
+    prompt = f"""Process the new GitHub {issue_type} below and return labels plus a first response.
 
 CONTEXT:
 - Repository: {repo_name}
 - Organization: {org_name}
-- Repository URL: https://github.com/{event.repository}
+- Repository metadata: {repository_context or "No additional metadata provided."}
 - User: {username}
 
-INSTRUCTIONS:
-- Do not answer the question or resolve the issue directly
-- Adapt the example {issue_type} response below as appropriate, keeping all badges, links and references provided
-- For bug reports, specifically request a minimum reproducible example (MRE) if not provided
-- INCLUDE ALL LINKS AND INSTRUCTIONS IN THE EXAMPLE BELOW, customized as appropriate
-- Mention to the user that this is an automated response and that an Ultralytics engineer will also assist soon
-- Do not add a sign-off or valediction like "best regards" at the end of your response
-- Do not add spaces between bullet points or numbered lists
-- Only link to files or URLs in the example below, do not add external links
-- Use a few emojis to enliven your response
+LABELS:
+- Select 0-3 labels only from the available names below; descriptions are authoritative
+- Use Alert only with high confidence for spam, abuse, or illegal content
+- Use bug only when the report describes concrete incorrect behavior with enough evidence to investigate; do not require a particular language, package manager, or reproduction format
+- Do not repeat a current label or guess from the repository name alone
+
+AVAILABLE LABELS:
+{labels_str}
+
+FIRST RESPONSE:
+- Acknowledge the concrete request or failure in the opening sentence without merely restating the title
+- Ask only for specific information that is materially missing; do not repeat a generic checklist or request details already supplied
+- For bugs, missing evidence may include minimal reproduction steps, observed and expected behavior, relevant environment/toolchain/dependency versions, and focused logs
+- For feature requests, ask at most one question that would materially change scope or behavior
+- Do not diagnose a cause or claim a resolution without evidence
+- Stay repository- and technology-aware; never assume Python, pip, PyPI, a branch name, or a release process
+- State that this is automated and an Ultralytics engineer will assist soon
+- {"Use the configured example as authoritative repository guidance; retain only instructions and links relevant to this report, and condense repetition" if configured_example else "Use at most 140 words"}
+- Do not add a heading, sign-off, or external links not present in the configured example
 
 EXAMPLE {issue_type.upper()} RESPONSE:
 {example}
@@ -167,7 +121,7 @@ EXAMPLE {issue_type.upper()} RESPONSE:
 {issue_type.upper()} DESCRIPTION:
 {body[:16000]}
 
-YOUR {issue_type.upper()} RESPONSE:
+Return the labels and final comment only.
 """
     messages = [
         {
@@ -176,7 +130,27 @@ YOUR {issue_type.upper()} RESPONSE:
         },
         {"role": "user", "content": prompt},
     ]
-    return get_response(messages)
+    schema = {
+        "type": "object",
+        "properties": {
+            "labels": {"type": "array", "items": {"type": "string"}},
+            "first_comment": {"type": "string"},
+        },
+        "required": ["labels", "first_comment"],
+        "additionalProperties": False,
+    }
+    response = get_response(
+        messages,
+        text_format={"format": {"type": "json_schema", "name": "first_interaction", "strict": True, "schema": schema}},
+    )
+    response["first_comment"] = check_links_in_string(
+        response.get("first_comment", "").replace(" @giscus[bot]", ""), replace=True
+    )
+    available = {name.lower(): name for name in filtered_labels}
+    response["labels"] = [
+        available[label.lower()] for label in response.get("labels", []) if label.lower() in available
+    ]
+    return response
 
 
 def main(*args, **kwargs):
@@ -186,32 +160,68 @@ def main(*args, **kwargs):
         return
 
     number, node_id, title, body, username, issue_type, action = get_event_content(event)
-    available_labels = event.get_repo_data("labels")
+    if issue_type == "pull request" and action == "opened" and event.should_skip_pr_author():
+        return
+    if issue_type != "pull request" and action not in {"opened", "created"}:
+        return
+
+    available_labels = (
+        event.paginate(f"{GITHUB_API_URL}/repos/{event.repository}/labels", hard=True) if AUTO_LABELS else []
+    )
     label_descriptions = {label["name"]: label.get("description") or "" for label in available_labels}
+    repository_data = event.event_data.get("repository", {})
+    repository_context = "; ".join(
+        str(value)
+        for value in (
+            repository_data.get("description"),
+            f"primary language: {repository_data['language']}" if repository_data.get("language") else None,
+            f"default branch: {repository_data['default_branch']}" if repository_data.get("default_branch") else None,
+        )
+        if value
+    )
 
     # Use unified PR open response for new PRs (summary + labels + first comment in 1 API call)
     if issue_type == "pull request" and action == "opened":
-        if event.should_skip_pr_author():
-            return
+        if AUTO_PR_SUMMARY or AUTO_LABELS:
+            print(f"Processing PR open by @{username} with unified API call...")
+            diff = event.get_pr_diff()
+            response = get_pr_open_response(
+                event.repository,
+                diff,
+                title,
+                username,
+                label_descriptions,
+                body,
+                repository_context,
+                summarize=AUTO_PR_SUMMARY,
+                acknowledge=AUTO_LABELS,
+                current_labels=[label["name"] for label in event.pr.get("labels", [])],
+            )
 
-        print(f"Processing PR open by @{username} with unified API call...")
-        diff = event.get_pr_diff()
-        response = get_pr_open_response(event.repository, diff, title, username, label_descriptions)
+            if AUTO_PR_SUMMARY and (summary := response.get("summary")):
+                print("Updating PR description with summary...")
+                skipped_dropdown = format_skipped_files_dropdown(response.get("skipped_files", []))
+                event.update_pr_description(
+                    number,
+                    f"{SUMMARY_MARKER}\n\n{ACTIONS_CREDIT}\n\n{summary}{skipped_dropdown}",
+                    fallback_description=event.pr.get("body") or "",
+                )
+                if sum(not char.isspace() for char in body) < 30:
+                    body = f"{body.rstrip()}\n\n{summary}".lstrip()
 
-        if summary := response.get("summary"):
-            print("Updating PR description with summary...")
-            skipped_dropdown = format_skipped_files_dropdown(response.get("skipped_files", []))
-            event.update_pr_description(number, f"{SUMMARY_MARKER}\n\n{ACTIONS_CREDIT}\n\n{summary}{skipped_dropdown}")
-            if sum(not char.isspace() for char in body) < 30:
-                body = f"{body.rstrip()}\n\n{summary}".lstrip()
-
-        if relevant_labels := response.get("labels", []):
-            apply_and_check_labels(event, number, node_id, issue_type, username, relevant_labels, label_descriptions)
-
-        if first_comment := response.get("first_comment"):
-            print("Adding first interaction comment...")
-            time.sleep(1)  # sleep to ensure label added first
-            event.add_comment(number, node_id, first_comment, issue_type)
+            if AUTO_LABELS:
+                apply_and_check_labels(
+                    event,
+                    number,
+                    node_id,
+                    issue_type,
+                    username,
+                    response.get("labels", []),
+                    label_descriptions,
+                )
+                if first_comment := response.get("first_comment"):
+                    print("Adding first interaction comment...")
+                    event.add_comment(number, node_id, first_comment, issue_type)
 
         # Automatic PR review after first interaction
         if AUTO_PR_REVIEW:
@@ -220,17 +230,13 @@ def main(*args, **kwargs):
         return
 
     # Handle issues and discussions (NOT PRs)
-    current_labels = (
-        []
-        if issue_type == "discussion"
-        else [label["name"].lower() for label in event.get_repo_data(f"issues/{number}/labels")]
+    item = event.event_data.get("issue", {}) if issue_type == "issue" else {}
+    current_labels = [label["name"].lower() for label in item.get("labels", [])]
+    response = get_first_interaction_response(
+        event, issue_type, title, body, username, label_descriptions, current_labels, repository_context
     )
-
-    relevant_labels = get_relevant_labels(issue_type, title, body, label_descriptions, current_labels)
-    apply_and_check_labels(event, number, node_id, issue_type, username, relevant_labels, label_descriptions)
-
-    if action in {"opened", "created"}:
-        custom_response = get_first_interaction_response(event, issue_type, title, body, username)
+    apply_and_check_labels(event, number, node_id, issue_type, username, response.get("labels", []), label_descriptions)
+    if custom_response := response.get("first_comment"):
         event.add_comment(number, node_id, custom_response, issue_type)
 
 
diff --git a/actions/review_pr.py b/actions/review_pr.py
index 1adb465..1381837 100644
--- a/actions/review_pr.py
+++ b/actions/review_pr.py
@@ -27,16 +27,17 @@ from .utils.openai_utils import _is_anthropic_model
 REVIEW_MARKER = "## 🔍 PR Review"
 ERROR_MARKER = "⚠️ Review generation encountered an error"
 EMOJI_MAP = {"CRITICAL": "❗", "HIGH": "⚠️", "MEDIUM": "💡", "LOW": "📝", "SUGGESTION": "💭"}
-MAX_CONTEXT_FILE_CHARS = 5000
+MAX_CONTEXT_FILE_CHARS = 12000
 MAX_REVIEW_COMMENTS = 8
-MAX_TOOL_OUTPUT_CHARS = 20000
-MAX_TOOL_FILE_LINES = 240
-MAX_AGENT_TURNS = 16
+MAX_TOOL_OUTPUT_CHARS = 40000
+MAX_TOOL_FILE_LINES = 400
+MAX_AGENT_TURNS = 24
+REVIEW_PROMPT_CHARS = round(MAX_PROMPT_CHARS * 1.5)  # reviews favor evidence depth over one-shot cost
 MAX_HISTORY_REVIEWS = 5  # prior reviews included in the prompt (the full history stays available via the tool)
 MAX_HISTORY_ITEM_CHARS = 8000  # per prior review or response, enough for a full summary plus its findings log
 MAX_HISTORY_CHARS = 20000
 MAX_FINDING_LOG_CHARS = 500  # per finding logged in the review body, the record that outlives its inline comment
-REVIEW_COST_SOFT_LIMIT = 2.00  # stop requesting tools after cumulative spend reaches this amount
+REVIEW_COST_SOFT_LIMIT = 5.00  # quality-first budget; stop requesting tools after this soft limit
 SEVERITY_RANK = {"CRITICAL": 0, "HIGH": 1, "MEDIUM": 2, "LOW": 3, "SUGGESTION": 4, None: 5}
 
 
@@ -535,7 +536,7 @@ def generate_pr_review(
                 snippet = f"{snippet.rstrip()}\n... (truncated)"
             # Only include if within budget, include buffer for Markdown noise
             estimated_cost = len(snippet) + 200
-            if total_chars + estimated_cost >= MAX_PROMPT_CHARS:
+            if total_chars + estimated_cost >= REVIEW_PROMPT_CHARS:
                 break  # Stop when we hit budget limit
             file_contents.append(f"### {file_path}\n```\n{snippet}\n```")
             total_chars += estimated_cost
@@ -543,7 +544,9 @@ def generate_pr_review(
             full_files_section = f"FULL FILE CONTENTS:\n{chr(10).join(file_contents)}\n\n"
 
     # Calculate remaining budget for diff and check if truncation needed
-    diff_budget = max(1000, MAX_PROMPT_CHARS - len(guidelines_section) - len(full_files_section) - len(history_section))
+    diff_budget = max(
+        1000, REVIEW_PROMPT_CHARS - len(guidelines_section) - len(full_files_section) - len(history_section)
+    )
     diff_truncated = len(augmented_diff) > diff_budget
     is_large_pr = diff_truncated or len(file_list) > 30
     if is_agent_review_model:  # must match the get_agent_response fallback gate
@@ -596,21 +599,23 @@ def generate_pr_review(
     content = (
         "You are an expert code reviewer for Ultralytics. Review code changes and provide inline comments ONLY for genuine issues.\n\n"
         "WHEN TO COMMENT (priority order):\n"
-        "- Bugs and logic errors that will cause failures\n"
-        "- Performance issues with measurable impact\n"
-        "- Code best practices and maintainability\n"
-        "- Missing error handling for likely failure cases\n"
-        "- Security issues (only obvious vulnerabilities, not speculative)\n\n"
+        "- Security vulnerabilities, data loss, corruption, or irreversible behavior\n"
+        "- Bugs, broken contracts, race conditions, and compatibility regressions with a concrete failure path\n"
+        "- Performance regressions with a plausible workload and measurable impact\n"
+        "- Maintainability only when the change creates demonstrated duplication, conflicting owners, or unreachable behavior\n\n"
         "WHEN NOT TO COMMENT:\n"
         "- Style/formatting (handled by ruff/prettier)\n"
         "- Minor naming preferences\n"
-        "- 'Consider using X' without clear benefit\n"
+        "- Generic best practices, defensive guards for impossible states, or 'consider using X' without a concrete defect\n"
+        "- Missing tests unless a high-risk regression path is uncovered and existing coverage does not exercise it\n"
         "- Issues in unchanged context lines\n\n"
         f"{visibility_section}"
         f"{continuity_section}"
         "QUALITY OVER QUANTITY:\n"
         "- Zero comments is valid for clean PRs: never invent an issue, never withhold an evidence-backed one\n"
-        "- Each comment must be actionable with clear reasoning\n"
+        "- Before commenting, trace the changed value or control flow through its owner, callers, and tests; use repository tools when the diff alone cannot prove the claim\n"
+        "- Each comment must identify the triggering scenario, resulting incorrect behavior, and smallest owner-level correction\n"
+        "- Severity: CRITICAL enables compromise or broad irreversible loss; HIGH breaks a core path or loses data; MEDIUM breaks a realistic edge path; LOW is a bounded defect; SUGGESTION is non-blocking\n"
         "- Combine related issues into one comment\n"
         f"- Hard cap: {MAX_REVIEW_COMMENTS} comments maximum\n\n"
         "SUGGESTIONS:\n"
@@ -619,7 +624,7 @@ def generate_pr_review(
         "- Single-line fixes only: provide 'suggestion' without 'start_line' to replace the line at 'line'\n"
         "- Match the exact indentation of the original code\n"
         "- Avoid triple backticks (```) in suggestions as they break Markdown formatting\n\n"
-        "SUMMARY: brief overall assessment of what's good and what needs attention; say so plainly when the PR is clean.\n\n"
+        "SUMMARY: state the reviewed scope, the behavioral verdict, and any remaining risk in concise prose; say LGTM plainly when the PR is clean.\n\n"
         "DIFF LINE FORMAT (how to read line numbers):\n"
         '  R  123 +code here      <- \'R\' means RIGHT (new file), number is 123, use {"line": 123, "side": "RIGHT"}\n'
         '  L   45 -code here      <- \'L\' means LEFT (old file), number is 45, use {"line": 45, "side": "LEFT"}\n'
@@ -642,7 +647,7 @@ def generate_pr_review(
             "content": (
                 f"Review this PR in https://github.com/{repository}:\n\n"
                 f"TITLE:\n{pr_title}\n\n"
-                f"BODY:\n{remove_html_comments(pr_description or '')[:1000]}\n\n"
+                f"BODY:\n{remove_html_comments(pr_description or '')[:8000]}\n\n"
                 f"{guidelines_section}"
                 f"{history_section}"
                 f"{full_files_section}"
diff --git a/actions/summarize_pr.py b/actions/summarize_pr.py
index 43d3303..851d861 100644
--- a/actions/summarize_pr.py
+++ b/actions/summarize_pr.py
@@ -2,8 +2,6 @@
 
 from __future__ import annotations
 
-from urllib.parse import urlparse
-
 from .utils import (
     ACTIONS_CREDIT,
     GITHUB_API_URL,
@@ -11,82 +9,15 @@ from .utils import (
     format_skipped_files_dropdown,
     get_pr_summary_prompt,
     get_response,
+    remove_html_comments,
 )
 
 SUMMARY_MARKER = "## 🛠️ PR Summary"
 
 
-def generate_merge_message(pr_summary, pr_credit, pr_url):
-    """Generates a motivating thank-you message for merged PR contributors."""
-    messages = [
-        {
-            "role": "system",
-            "content": (
-                "You are an Ultralytics AI assistant. Your response is posted verbatim as a GitHub "
-                "comment on a merged PR. Respond with ONLY the final comment body, ready to post. "
-                "Do not include preambles (e.g. 'Absolutely', 'Sure', 'Here's a comment'), "
-                "explanations, sign-offs, or horizontal-rule separators ('---')."
-            ),
-        },
-        {
-            "role": "user",
-            "content": (
-                f"Compose the thank-you comment for the merged PR {pr_url} by {pr_credit}. "
-                f"Context:\n{pr_summary}\n\n"
-                f"Start with an enthusiastic note about the merge, incorporate a relevant inspirational quote from a historical "
-                f"figure, and connect it to the PR's impact. Keep it concise yet meaningful, ensuring contributors feel valued."
-            ),
-        },
-    ]
-    return get_response(messages)
-
-
-def generate_issue_comment(pr_url, pr_summary, pr_credit, pr_title=""):
-    """Generates personalized issue comment based on PR context."""
-    # pr_url is the GraphQL PullRequest.url HTML URL (https://github.com/owner/repo/pull/N) in production;
-    # REST API URLs (https://api.github.com/repos/owner/repo/pulls/N) are also supported
-    parsed = urlparse(pr_url)
-    if "/repos/" in pr_url and "/pulls/" in pr_url:
-        repo_parts = pr_url.split("/repos/", 1)[1].split("/pulls/", 1)[0]
-    elif parsed.hostname == "github.com" and "/pull/" in parsed.path:
-        repo_parts = parsed.path.lstrip("/").split("/pull/", 1)[0]
-    else:
-        repo_parts = ""
-    owner_repo = repo_parts.split("/")
-    repo_name = owner_repo[-1] if len(owner_repo) > 1 else "package"
-
-    messages = [
-        {
-            "role": "system",
-            "content": (
-                "You are an Ultralytics AI assistant. Your response is posted verbatim as a GitHub "
-                "issue comment. Respond with ONLY the final comment body, ready to post. Do not "
-                "include preambles (e.g. 'Absolutely', 'Sure', 'Here's a comment'), explanations, "
-                "sign-offs, or horizontal-rule separators ('---'). No @ mentions or direct addressing."
-            ),
-        },
-        {
-            "role": "user",
-            "content": f"Compose a GitHub issue comment announcing a potential fix for this issue is now merged in linked PR {pr_url} by {pr_credit}\n\n"
-            f"PR Title: {pr_title}\n\n"
-            f"Context from PR:\n{pr_summary}\n\n"
-            f"Include:\n"
-            f"1. An explanation of key changes from the PR that may resolve this issue\n"
-            f"2. Credit to the PR author and contributors\n"
-            f"3. Options for testing if PR changes have resolved this issue:\n"
-            f"   - If the PR mentions a specific version number (like v8.0.0 or 3.1.0), include: pip install -U {repo_name}>=VERSION\n"
-            f"   - Also suggest: pip install git+https://github.com/{repo_parts}.git@main\n"
-            f"   - If appropriate, mention they can also wait for the next official PyPI release\n"
-            f"4. Request feedback on whether the PR changes resolve the issue\n"
-            f"5. Thank 🙏 for reporting the issue and welcome any further feedback if the issue persists\n\n",
-        },
-    ]
-    return get_response(messages)
-
-
-def generate_pr_summary(repository, diff_text):
+def generate_pr_summary(repository, diff_text, title="", description=""):
     """Generates a concise, professional summary of a PR using the OpenAI or Anthropic API."""
-    prompt, is_large, skipped_files = get_pr_summary_prompt(repository, diff_text)
+    prompt, is_large, skipped_files = get_pr_summary_prompt(repository, diff_text, title, description)
 
     messages = [
         {
@@ -108,10 +39,20 @@ def generate_pr_summary(repository, diff_text):
 def label_fixed_issues(event, pr_summary):
     """Labels issues closed by PR when merged, notifies users, and returns PR contributors."""
     pr_credit, data = event.get_pr_contributors()
-    if not pr_credit:
+    if not data:
         return None
 
-    comment = generate_issue_comment(data["url"], pr_summary, pr_credit, data.get("title") or "")
+    credit = f" by {pr_credit}" if pr_credit else ""
+    synopsis = ""
+    if "### 🌟 Summary" in pr_summary and "### 📊 Key Changes" in pr_summary:
+        synopsis = pr_summary.split("### 🌟 Summary", 1)[1].split("### 📊 Key Changes", 1)[0].strip()
+    details = f"\n\n{synopsis}" if synopsis else ""
+    comment = (
+        f"A potential fix is now available in the [merged pull request]({data['url']}){credit}.{details}\n\n"
+        "Please test the merged change using "
+        "this repository's documented workflow. If the issue persists, "
+        "share the updated behavior and any new diagnostic details. Thank you for reporting it! 🙏"
+    )
 
     for issue in data["closingIssuesReferences"]["nodes"]:
         number = issue["number"]
@@ -136,11 +77,14 @@ def main(*args, **kwargs):
 
     # Generate PR summary
     print("Generating PR summary...")
-    summary = generate_pr_summary(event.repository, diff)
+    description = (event.pr.get("body") or "").split(SUMMARY_MARKER)[0]
+    summary = generate_pr_summary(
+        event.repository, diff, event.pr.get("title") or "", remove_html_comments(description)
+    )
 
     # Update PR description
     print("Updating PR description...")
-    event.update_pr_description(event.pr["number"], summary)
+    event.update_pr_description(event.pr["number"], summary, fallback_description=event.pr.get("body") or "")
 
     if event.pr.get("merged"):
         print("PR is merged, labeling fixed issues...")
@@ -149,9 +93,9 @@ def main(*args, **kwargs):
         event.remove_labels(event.pr["number"], labels=("TODO",))
         if pr_credit:
             print("Posting PR author thank you message...")
-            pr_url = f"{GITHUB_API_URL}/repos/{event.repository}/pulls/{event.pr['number']}"
-            message = generate_merge_message(summary, pr_credit, pr_url)
-            event.add_comment(event.pr["number"], None, message, "pull request")
+            event.add_comment(
+                event.pr["number"], None, f"🚀 Merged! Thank you {pr_credit} for the contribution.", "pull request"
+            )
 
 
 if __name__ == "__main__":
diff --git a/actions/utils/github_utils.py b/actions/utils/github_utils.py
index 289440b..3af2c6f 100644
--- a/actions/utils/github_utils.py
+++ b/actions/utils/github_utils.py
@@ -312,19 +312,20 @@ class Action:
             print(result.get("errors"))
         return result
 
-    def update_pr_description(self, number: int, new_summary: str, max_retries: int = 2):
-        """Updates PR description with summary, retrying if description is None."""
+    def update_pr_description(
+        self, number: int, new_summary: str, max_retries: int = 2, fallback_description: str = ""
+    ):
+        """Update a PR description with a summary while preserving its current body."""
         import time
 
         url = f"{GITHUB_API_URL}/repos/{self.repository}/pulls/{number}"
-        description = ""
-        for i in range(max_retries + 1):
+        description = self.get(url).json().get("body") or ""
+        for _ in range(max_retries if not description and not fallback_description else 0):
+            time.sleep(1)
             description = self.get(url).json().get("body") or ""
             if description:
                 break
-            if i < max_retries:
-                print("No current PR description found, retrying...")
-                time.sleep(1)
+        description = description or fallback_description
 
         start = "## 🛠️ PR Summary"
         if start in description:
@@ -347,7 +348,7 @@ class Action:
 
     def apply_labels(self, number: int, node_id: str, labels: list[str], issue_type: str):
         """Applies specified labels to a GitHub issue, pull request, or discussion."""
-        if "Alert" in labels:
+        if any(label.lower() == "alert" for label in labels):
             self.create_alert_label()
 
         if issue_type == "discussion":
diff --git a/actions/utils/openai_utils.py b/actions/utils/openai_utils.py
index 73bfe99..cfc5570 100644
--- a/actions/utils/openai_utils.py
+++ b/actions/utils/openai_utils.py
@@ -63,9 +63,9 @@ SYSTEM_PROMPT_ADDITION = """Guidance:
   - Markdown: Reply in Markdown format.
   - Links: Use descriptive anchor text for all URLs.
   - Code:
-    - Provide minimal code examples if helpful.
-    - Enclose code in backticks: `pip install ultralytics` for inline code or e.g. ```python for larger code blocks.
-    - Think and verify the argument names, methods, class and files used in your code examples for accuracy.
+    - Provide minimal code examples only when supported by the repository context.
+    - Use inline backticks or a fenced block with the repository's appropriate language.
+    - Verify argument names, methods, classes, files, package managers, and commands before including them.
   - Use the "@" symbol to refer to GitHub users, e.g. @glenn-jocher.
   - Tone: Adopt a professional, friendly, and concise tone.
 """
@@ -89,28 +89,30 @@ def remove_outer_codeblocks(string):
 
 def filter_labels(available_labels: dict, current_labels: list | None = None, is_pr: bool = False) -> dict:
     """Filters labels by removing manually-assigned and mutually exclusive labels, adding an Alert label if absent."""
-    current_labels = current_labels or []
-    filtered = available_labels.copy()
-
-    for label in (
+    current_labels = {label.lower() for label in (current_labels or [])}
+    excluded = {
         "help wanted",
-        "TODO",
+        "todo",
         "research",
         "non-reproducible",
         "popular",
         "invalid",
-        "Stale",
+        "stale",
         "wontfix",
         "duplicate",
-    ):
-        filtered.pop(label, None)
+    }
+    filtered = {
+        name: description
+        for name, description in available_labels.items()
+        if name.lower() not in excluded | current_labels
+    }
 
     if "bug" in current_labels:
-        filtered.pop("question", None)
+        filtered = {name: description for name, description in filtered.items() if name.lower() != "question"}
     elif "question" in current_labels:
-        filtered.pop("bug", None)
+        filtered = {name: description for name, description in filtered.items() if name.lower() != "bug"}
 
-    if "Alert" not in filtered:
+    if "alert" not in current_labels and not any(name.lower() == "alert" for name in filtered):
         filtered["Alert"] = (
             "Potential spam, abuse, or illegal activity including advertising, unsolicited promotions, malware, "
             "phishing, crypto offers, pirated software or media, free movie downloads, cracks, keygens or any other "
@@ -122,39 +124,34 @@ def filter_labels(available_labels: dict, current_labels: list | None = None, is
 
 def get_pr_summary_guidelines() -> str:
     """Returns PR summary formatting guidelines (used by both unified PR open and PR update/merge)."""
-    return """Summarize this PR, focusing on major changes, their purpose, and potential impact. Keep the summary clear and concise, suitable for a broad audience. Add emojis to enliven the summary. Your response must include all 3 sections below with their H3 Markdown headers (do not use H1 or H2 headers):
+    return """Summarize the implemented change, not the author's intentions. Ground every statement in the diff and PR context; never invent behavior, test results, compatibility claims, or user impact. Name the concrete components and behaviors changed, consolidate related edits, and omit file-by-file narration and generic praise. Use concise plain language for maintainers and users. Your response must include exactly these 3 sections with H3 Markdown headers (do not use H1 or H2 headers):
 
 ### 🌟 Summary
-(single-line synopsis)
+(one sentence stating what changed and why)
 
 ### 📊 Key Changes
-- (bullet points highlighting major changes)
+- (2-5 specific bullets, ordered by importance)
 
 ### 🎯 Purpose & Impact
-- (bullet points explaining benefits and potential impact to users)"""
+- (specific behavior or workflow impact; say "No user-facing change" when appropriate)"""
 
 
-def get_pr_summary_prompt(repository: str, diff_text: str) -> tuple[str, bool, list[str]]:
+def get_pr_summary_prompt(
+    repository: str, diff_text: str, title: str = "", description: str = ""
+) -> tuple[str, bool, list[str]]:
     """Returns the complete PR summary generation prompt with filtered diff (used by PR update/merge)."""
     filtered_diff, skipped_files = filter_diff_text(diff_text)
-    prompt = f"{get_pr_summary_guidelines()}\n\nRepository: '{repository}'\n\nHere's the PR diff:\n\n{filtered_diff[:MAX_PROMPT_CHARS]}"
+    prompt = (
+        f"{get_pr_summary_guidelines()}\n\nRepository: {repository}\nPR title: {title}\n"
+        f"PR description:\n{description[:8000]}\n\nPR diff:\n{filtered_diff[:MAX_PROMPT_CHARS]}"
+    )
     prompt += format_skipped_files_note(skipped_files)
     return prompt, len(filtered_diff) > MAX_PROMPT_CHARS, skipped_files
 
 
 def get_pr_first_comment_template(repository: str, username: str) -> str:
-    """Returns the PR first comment template with checklist (used only by unified PR open)."""
-    return f"""👋 Hello @{username}, thank you for submitting a `{repository}` 🚀 PR! To ensure a seamless integration of your work, please review the following checklist:
-
-- ✅ **Define a Purpose**: Clearly explain the purpose of your fix or feature in your PR description, and link to any [relevant issues](https://github.com/{repository}/issues). Ensure your commit messages are clear, concise, and adhere to the project's conventions.
-- ✅ **Synchronize with Source**: Confirm your PR is synchronized with the `{repository}` `main` branch. If it's behind, update it by clicking the 'Update branch' button or by running `git pull` and `git merge main` locally.
-- ✅ **Ensure CI Checks Pass**: Verify all Ultralytics [Continuous Integration (CI)](https://docs.ultralytics.com/help/CI) checks are passing. If any checks fail, please address the issues.
-- ✅ **Update Documentation**: Update the relevant [documentation](https://docs.ultralytics.com/) for any new or modified features.
-- ✅ **Add Tests**: If applicable, include or update tests to cover your changes, and confirm that all tests are passing.
-- ✅ **Sign the CLA**: Please ensure you have signed our [Contributor License Agreement](https://docs.ultralytics.com/help/CLA) if this is your first Ultralytics PR by writing "I have read the CLA Document and I sign the CLA" in a new message.
-- ✅ **Minimize Changes**: Limit your changes to the **minimum** necessary for your bug fix or feature addition. _"It is not daily increase but daily decrease, hack away the unessential. The closer to the source, the less wastage there is."_  — Bruce Lee
-
-For more guidance, please refer to our [Contributing Guide](https://docs.ultralytics.com/help/contributing). Don't hesitate to leave a comment if you have any questions. Thank you for contributing to Ultralytics! 🚀"""
+    """Return the concise acknowledgment that the PR-open model customizes."""
+    return f"👋 Thanks @{username} for the `{repository}` contribution! This is an automated first response; an Ultralytics engineer will review the PR soon."
 
 
 def _is_anthropic_model(model: str) -> bool:
@@ -662,46 +659,69 @@ def get_response(
             raise
 
 
-def get_pr_open_response(repository: str, diff_text: str, title: str, username: str, available_labels: dict) -> dict:
+def get_pr_open_response(
+    repository: str,
+    diff_text: str,
+    title: str,
+    username: str,
+    available_labels: dict,
+    description: str = "",
+    repository_context: str = "",
+    summarize: bool = True,
+    acknowledge: bool = True,
+    current_labels: list | None = None,
+) -> dict:
     """Generates unified PR response with summary, labels, and first comment in a single API call."""
     filtered_diff, skipped_files = filter_diff_text(diff_text)
     is_large = len(filtered_diff) > MAX_PROMPT_CHARS
 
-    filtered_labels = filter_labels(available_labels, is_pr=True)
+    filtered_labels = filter_labels(available_labels, current_labels) if available_labels else {}
     labels_str = "\n".join(f"- {name}: {description}" for name, description in filtered_labels.items())
+    summary_instructions = get_pr_summary_guidelines() if summarize else "Return an empty summary string."
+    label_instructions = (
+        "Array of 1-3 most relevant label names. Only use Alert with high confidence for inappropriate PRs. "
+        f"Return an empty array if no label is relevant. Available labels:\n{labels_str}"
+        if filtered_labels
+        else "Return an empty labels array."
+    )
+    comment_instructions = (
+        f"""Write a concise acknowledgment that:
+- Opens with one specific sentence showing what the PR changes; do not repeat the generated summary
+- Requests an action only when the description or diff provides concrete evidence it is missing
+- Never claims tests passed, asks for generic checklist work, or assumes a language, package manager, branch name, or release process
+- States that the response is automated and an Ultralytics engineer will review soon
+- Uses at most 90 words, no headings, sign-off, inspirational quote, or external links
+
+Base acknowledgment:
+{get_pr_first_comment_template(repository, username)}"""
+        if acknowledge
+        else "Return an empty first_comment string."
+    )
 
     prompt = f"""You are processing a new GitHub PR by @{username} for the {repository} repository.
 
+Repository context: {repository_context or "No additional metadata provided."}
+PR description: {description[:8000] or "No description provided."}
+
 Generate 3 outputs in a single JSON response for the PR titled '{title}' with the following diff:
 {filtered_diff[:MAX_PROMPT_CHARS]}{format_skipped_files_note(skipped_files)}
 
 
 --- FIRST JSON OUTPUT (PR SUMMARY) ---
-{get_pr_summary_guidelines()}
+{summary_instructions}
 
 --- SECOND JSON OUTPUT (PR LABELS) ---
-Array of 1-3 most relevant label names. Only use "Alert" with high confidence for inappropriate PRs. Return empty array if no labels relevant. Available labels:
-{labels_str}
+{label_instructions}
 
---- THIRD OUTPUT (PR FIRST COMMENT) ---
-Customized welcome message adapting the template below:
-- INCLUDE ALL LINKS AND INSTRUCTIONS from the template below, customized as appropriate
-- Keep all checklist items and links from template
-- Only link to files or URLs in the template below, do not add external links
-- Mention this is automated and an engineer will assist
-- Use a few emojis
-- No sign-off or "best regards"
-- No spaces between bullet points
-
-Example comment template (adapt as needed, keep all links):
-{get_pr_first_comment_template(repository, username)}"""
+--- THIRD JSON OUTPUT (PR FIRST COMMENT) ---
+{comment_instructions}"""
 
     schema = {
         "type": "object",
         "properties": {
             "summary": {"type": "string", "description": "PR summary with emoji sections"},
             "labels": {"type": "array", "items": {"type": "string"}, "description": "Array of label names"},
-            "first_comment": {"type": "string", "description": "Welcome comment with checklist"},
+            "first_comment": {"type": "string", "description": "Concise automated acknowledgment"},
         },
         "required": ["summary", "labels", "first_comment"],
         "additionalProperties": False,
diff --git a/dependabot/action.yml b/dependabot/action.yml
index 2c7eeed..aca70da 100644
--- a/dependabot/action.yml
+++ b/dependabot/action.yml
@@ -33,20 +33,15 @@ runs:
 
     - name: Install ultralytics-actions
       env:
-        GITHUB_REPOSITORY: ${{ github.repository }}
-        GITHUB_SHA: ${{ github.sha }}
+        ACTION_PATH: ${{ github.action_path }}/..
       run: |
         echo "::group::Install ultralytics-actions"
-        if [ "$GITHUB_REPOSITORY" = "ultralytics/actions" ]; then
-          echo "Installing from commit: $GITHUB_SHA"
-          packages="https://github.com/ultralytics/actions/archive/${GITHUB_SHA}.tar.gz"
-        else
-          packages="https://github.com/ultralytics/actions/archive/refs/heads/main.tar.gz"
-        fi
+        package_dir=$(mktemp -d "$RUNNER_TEMP/ultralytics-actions.XXXXXX")
+        uv build --wheel "$ACTION_PATH" --out-dir "$package_dir"
         if [ "$(uname)" = "Darwin" ]; then
-          uv pip install --system --break-system-packages $packages
+          uv pip install --system --break-system-packages "$package_dir"/*.whl
         else
-          sudo env "PATH=$PATH" uv pip install --system --break-system-packages $packages
+          sudo env "PATH=$PATH" uv pip install --system --break-system-packages "$package_dir"/*.whl
         fi
         echo "::endgroup::"
       shell: bash
diff --git a/github-report/action.yml b/github-report/action.yml
index fed3a59..b4b37d9 100644
--- a/github-report/action.yml
+++ b/github-report/action.yml
@@ -49,20 +49,15 @@ runs:
 
     - name: Install ultralytics-actions
       env:
-        GITHUB_REPOSITORY: ${{ github.repository }}
-        GITHUB_SHA: ${{ github.sha }}
+        ACTION_PATH: ${{ github.action_path }}/..
       run: |
         echo "::group::Install ultralytics-actions"
-        if [ "$GITHUB_REPOSITORY" = "ultralytics/actions" ]; then
-          echo "Installing from commit: $GITHUB_SHA"
-          packages="https://github.com/ultralytics/actions/archive/${GITHUB_SHA}.tar.gz"
-        else
-          packages="https://github.com/ultralytics/actions/archive/refs/heads/main.tar.gz"
-        fi
+        package_dir=$(mktemp -d "$RUNNER_TEMP/ultralytics-actions.XXXXXX")
+        uv build --wheel "$ACTION_PATH" --out-dir "$package_dir"
         if [ "$(uname)" = "Darwin" ]; then
-          uv pip install --system --break-system-packages $packages
+          uv pip install --system --break-system-packages "$package_dir"/*.whl
         else
-          sudo env "PATH=$PATH" uv pip install --system --break-system-packages $packages
+          sudo env "PATH=$PATH" uv pip install --system --break-system-packages "$package_dir"/*.whl
         fi
         echo "::endgroup::"
       shell: bash
diff --git a/setup-uv/README.md b/setup-uv/README.md
index 7c15094..0faa867 100644
--- a/setup-uv/README.md
+++ b/setup-uv/README.md
@@ -1,6 +1,6 @@
 # Setup uv Action
 
-Installs the latest [uv](https://docs.astral.sh/uv/) release with retry support and optionally activates a Python virtual environment.
+Provides Ultralytics defaults for the official [uv setup action](https://github.com/astral-sh/setup-uv), with optional Python environment activation and dependency caching.
 
 ## Usage
 
diff --git a/setup-uv/action.yml b/setup-uv/action.yml
index 044c7ab..eb0c313 100644
--- a/setup-uv/action.yml
+++ b/setup-uv/action.yml
@@ -1,7 +1,7 @@
 # Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
 
 name: "Setup uv"
-description: "Install the latest uv with retry and optionally activate a Python environment"
+description: "Install uv with optional Python environment activation and dependency caching"
 inputs:
   python-version:
     description: "Python version for uv commands"
@@ -23,51 +23,10 @@ inputs:
 runs:
   using: "composite"
   steps:
-    - name: Install uv
-      uses: ultralytics/actions/retry@main
+    - uses: astral-sh/setup-uv@v9.0.0
       with:
-        run: |
-          set -o pipefail
-          if [ "$RUNNER_OS" = "Windows" ]; then
-            powershell -NoProfile -ExecutionPolicy Bypass -Command '$env:UV_UNMANAGED_INSTALL = $env:RUNNER_TEMP + "\uv-bin"; irm https://astral.sh/uv/install.ps1 | iex'
-            echo "$RUNNER_TEMP\uv-bin" >> "$GITHUB_PATH"
-          else
-            curl -LsSf https://astral.sh/uv/install.sh | env UV_UNMANAGED_INSTALL="$RUNNER_TEMP/uv-bin" sh
-            echo "$RUNNER_TEMP/uv-bin" >> "$GITHUB_PATH"
-          fi
-
-    - name: Configure environment
-      shell: bash
-      env:
-        ACTIVATE_ENVIRONMENT: ${{ inputs.activate-environment }}
-        INPUT_PYTHON_VERSION: ${{ inputs.python-version }}
-      run: |
-        if [ -n "$INPUT_PYTHON_VERSION" ]; then
-          export UV_PYTHON="$INPUT_PYTHON_VERSION"
-          echo "UV_PYTHON=$INPUT_PYTHON_VERSION" >> "$GITHUB_ENV"
-        fi
-        if [ "$ACTIVATE_ENVIRONMENT" = "true" ]; then
-          uv venv --clear
-          if [ "$RUNNER_OS" = "Windows" ]; then
-            venv_path=$(cygpath -w "$PWD/.venv")
-            echo "VIRTUAL_ENV=$venv_path" >> "$GITHUB_ENV"
-            echo "$venv_path\Scripts" >> "$GITHUB_PATH"
-          else
-            echo "VIRTUAL_ENV=$PWD/.venv" >> "$GITHUB_ENV"
-            echo "$PWD/.venv/bin" >> "$GITHUB_PATH"
-          fi
-        fi
-
-    - name: Find uv cache
-      if: inputs.enable-cache == 'true'
-      id: uv-cache
-      shell: bash
-      run: echo "path=$(uv cache dir)" >> "$GITHUB_OUTPUT"
-
-    - name: Cache uv
-      if: inputs.enable-cache == 'true'
-      uses: actions/cache@v6
-      with:
-        path: ${{ steps.uv-cache.outputs.path }}
-        key: uv-${{ runner.os }}-${{ runner.arch }}-${{ inputs.python-version }}-${{ hashFiles(inputs.cache-dependency-glob) }}
-        restore-keys: uv-${{ runner.os }}-${{ runner.arch }}-${{ inputs.python-version }}-
+        python-version: ${{ inputs.python-version }}
+        activate-environment: ${{ inputs.activate-environment }}
+        enable-cache: ${{ inputs.enable-cache }}
+        cache-dependency-glob: ${{ inputs.cache-dependency-glob }}
+        ignore-nothing-to-cache: true
diff --git a/tests/test_first_interaction.py b/tests/test_first_interaction.py
index c90f9eb..44cffc1 100644
--- a/tests/test_first_interaction.py
+++ b/tests/test_first_interaction.py
@@ -5,7 +5,7 @@ from unittest.mock import MagicMock, call, patch
 import pytest
 
 from actions import first_interaction, review_pr
-from actions.first_interaction import get_event_content, get_first_interaction_response, get_relevant_labels
+from actions.first_interaction import get_event_content, get_first_interaction_response
 
 
 def test_get_event_content_issue():
@@ -40,15 +40,15 @@ def test_get_event_content_pr():
     # Create mock event
     mock_event = MagicMock()
     mock_event.event_name = "pull_request"
-    mock_event.event_data = {"action": "opened", "pull_request": {"number": 456}}
-
-    # Mock PR data returned from API
-    mock_event.get_repo_data.return_value = {
-        "number": 456,
-        "node_id": "node456",
-        "title": "Test PR",
-        "body": "PR description",
-        "user": {"login": "testuser"},
+    mock_event.event_data = {
+        "action": "opened",
+        "pull_request": {
+            "number": 456,
+            "node_id": "node456",
+            "title": "Test PR",
+            "body": "PR description",
+            "user": {"login": "testuser"},
+        },
     }
 
     number, node_id, title, body, username, issue_type, action = get_event_content(mock_event)
@@ -60,6 +60,7 @@ def test_get_event_content_pr():
     assert username == "testuser"
     assert issue_type == "pull request"
     assert action == "opened"
+    mock_event.get_repo_data.assert_not_called()
 
 
 @patch("actions.first_interaction.review_pr.run_review")
@@ -86,7 +87,8 @@ def test_open_pr_review_description(mock_action, mock_content, mock_response, mo
     event = mock_action.return_value
     event.should_skip_llm.return_value = False
     event.should_skip_pr_author.return_value = False
-    event.get_repo_data.return_value = []
+    event.paginate.return_value = []
+    event.pr = {"body": body}
     event.get_pr_diff.return_value = "diff"
     event.get_pr_diff_snapshot.return_value = ("review diff", "headsha")
 
@@ -95,43 +97,36 @@ def test_open_pr_review_description(mock_action, mock_content, mock_response, mo
     mock_review.assert_called_once_with(event, "Test PR", expected)
 
 
+@patch("actions.first_interaction.check_links_in_string", return_value="Thank you for your issue")
 @patch("actions.first_interaction.get_response")
-def test_get_relevant_labels(mock_get_response):
-    """Test getting relevant labels for an issue."""
-    mock_get_response.return_value = "bug, enhancement"
-
-    available_labels = {
-        "bug": "A bug in the software",
-        "enhancement": "New feature or request",
-        "documentation": "Improvements to documentation",
+def test_get_first_interaction_response(mock_get_response, mock_check_links):
+    """Test labels and first response share one structured model call."""
+    mock_get_response.return_value = {
+        "labels": ["BUG", "not-a-label"],
+        "first_comment": "Thank you for your issue",
     }
 
-    labels = get_relevant_labels(
-        issue_type="issue",
-        title="Bug: App crashes on startup",
-        body="The application crashes when started",
-        available_labels=available_labels,
-        current_labels=[],
-    )
-
-    assert labels == ["bug", "enhancement"]
-    mock_get_response.assert_called_once()
-
-
-@patch("actions.first_interaction.get_response")
-def test_get_first_interaction_response(mock_get_response):
-    """Test generating first interaction response."""
-    mock_get_response.return_value = "Thank you for your issue"
-
     mock_event = MagicMock()
     mock_event.repository = "test/repo"
 
     response = get_first_interaction_response(
-        event=mock_event, issue_type="issue", title="Test Issue", body="Issue description", username="testuser"
+        event=mock_event,
+        issue_type="issue",
+        title="Test Issue",
+        body="Issue description",
+        username="testuser",
+        available_labels={"bug": "Something is broken"},
+        current_labels=[],
+        repository_context="A test repository; primary language: Rust",
     )
 
-    assert response == "Thank you for your issue"
+    assert response == {"labels": ["bug"], "first_comment": "Thank you for your issue"}
     mock_get_response.assert_called_once()
+    mock_check_links.assert_called_once_with("Thank you for your issue", replace=True)
+    assert mock_get_response.call_args.kwargs["text_format"]["format"]["type"] == "json_schema"
+    prompt = mock_get_response.call_args.args[0][1]["content"]
+    assert "primary language: Rust" in prompt
+    assert "never assume Python, pip, PyPI" in prompt
 
 
 @patch("actions.review_pr._verified_local_checkout", return_value=True)
diff --git a/tests/test_summarize_pr.py b/tests/test_summarize_pr.py
index 751a4e2..ea294c3 100644
--- a/tests/test_summarize_pr.py
+++ b/tests/test_summarize_pr.py
@@ -1,8 +1,8 @@
 # Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
 
-from unittest.mock import patch
+from unittest.mock import MagicMock, patch
 
-from actions.summarize_pr import generate_issue_comment, generate_merge_message, generate_pr_summary
+from actions.summarize_pr import generate_pr_summary, label_fixed_issues
 
 
 @patch("actions.summarize_pr.get_response")
@@ -16,35 +16,22 @@ def test_generate_pr_summary(mock_get_response):
     mock_get_response.assert_called_once()
 
 
-@patch("actions.summarize_pr.get_response")
-def test_generate_merge_message(mock_get_response):
-    """Test generating merge thank you messages."""
-    mock_get_response.return_value = "Thank you for your contribution!"
-
-    message = generate_merge_message(
-        pr_summary="Feature implementation", pr_credit="@testuser", pr_url="https://github.com/test/repo/pull/1"
+def test_label_fixed_issues_posts_repository_neutral_comment():
+    """Test merged-issue responses avoid language, package-manager, and branch assumptions."""
+    event = MagicMock(repository="owner/swift-app")
+    event.get_pr_contributors.return_value = (
+        "@testuser",
+        {
+            "url": "https://github.com/owner/swift-app/pull/123",
+            "title": "Fix launch crash",
+            "closingIssuesReferences": {"nodes": [{"number": 7}]},
+        },
     )
 
-    assert message == "Thank you for your contribution!"
-    mock_get_response.assert_called_once()
-
-
-@patch("actions.summarize_pr.get_response")
-def test_generate_issue_comment(mock_get_response):
-    """Test generating issue comments about PR fixes."""
-    mock_get_response.return_value = "This issue is fixed in PR #123"
-
-    for pr_url in ("https://github.com/owner/repo/pull/123", "https://api.github.com/repos/owner/repo/pulls/123"):
-        comment = generate_issue_comment(
-            pr_url=pr_url,
-            pr_summary="Fixed bug",
-            pr_credit="@testuser",
-            pr_title="Bug fix PR",
-        )
-
-        assert comment == "This issue is fixed in PR #123"
-        prompt = mock_get_response.call_args[0][0][1]["content"]
-        assert "pip install -U repo>=VERSION" in prompt
-        assert "pip install git+https://github.com/owner/repo.git@main" in prompt
-
-    assert mock_get_response.call_count == 2
+    summary = "## 🛠️ PR Summary\n\n### 🌟 Summary\nFixed launch handling.\n\n### 📊 Key Changes\n- Safer startup"
+    assert label_fixed_issues(event, summary) == "@testuser"
+    comment = event.post.call_args_list[1].kwargs["json"]["body"]
+    assert "[merged pull request](https://github.com/owner/swift-app/pull/123)" in comment
+    assert "documented workflow" in comment
+    assert "Fixed launch handling." in comment
+    assert all(term not in comment for term in ("pip", "PyPI", "@main"))