agentleFS
MrCipherSmithMrCipherSmith/goodai-base/skills/pr-issue-documenter/SKILL.md

pr-issue-documenter

Skill0 starsChanged 5 months agoMITView on GitHub

Use when documenting PR changes, adding a PR description, creating a linked issue for a PR, or updating an existing issue body.

The file

Show its rules and text

We split this skill into the rules it gives an agent

18 rules. About 0 are common, 14 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈14

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Check PR body for Closes #N, Fixes #N, Parent issue: #N

    line 118
  2. Very specific:

    Check issue body for Parent issue: #N

    line 119
  3. Distinctive:

    If parent issue found, fetch its body too:

    line 120
  4. Very specific:

    Use backticks for code identifiers (ClassName, methodName, fileName.ts)

    line 192
  5. Very specific:

    Use heredoc (cat <<'EOF') for body content to preserve formatting

    line 304
  6. Distinctive:

    Always verify the result after each operation

    line 305
  7. Distinctive:

    After all operations, report to the user:

    line 309
  8. Distinctive:

    When comparing existing issue description with PR diff:

    line 321
  9. Distinctive:

    Always present contradictions to user before making changes.

    line 329
  10. Distinctive:

    Always use English for code documentation

    line 335
  11. Distinctive:

    Never invent changes not present in the diff

    line 336
  12. Distinctive:

    Always verify diff content before writing descriptions

    line 337
  13. Distinctive:

    DO check for contradictions before updating existing descriptions

    line 356
  14. Distinctive:

    DO ask user before overwriting existing issue content

    line 357
  15. Distinctive:

    DO update parent issue when creating sub-issues

    line 358
  16. Distinctive:

    DO NOT: create issues without user confirmation

    line 362
  17. Distinctive:

    DO NOT: write comments on GitHub PRs/issues (only edit body)

    line 364
  18. Distinctive:

    If called within an orchestrator job context, check for job context before starting:

    line 368

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

PR & Issue Documenter

Purpose

Analyzes PR commits and diffs to generate structured, accurate descriptions for GitHub PRs and linked issues. Handles the full lifecycle: PR description, sub-issue creation, parent issue updates, and contradiction detection with existing descriptions.

Input: PR URL, issue URL, and/or commit SHAs Output: Updated PR body, created/updated issue body, parent issue link

When to Use

  • User provides PR link and asks for a description
  • User asks to document changes in a PR
  • User wants to create an issue for a PR
  • User wants to update an existing issue with PR changes
  • User provides commit SHAs and wants documentation

Architecture

Step 1: Parse Input
    |
Step 2: Collect Context (commits, diff, existing descriptions)
    |
Step 3: Analyze Changes (categorize, group, identify patterns)
    |
Step 4: Generate PR Description
    |
Step 5: Handle Issue (create / update / skip)
    |
Step 6: Apply Changes (gh pr edit, gh issue edit/create)
    |
Step 7: Verify & Report

Workflow

Copy this checklist and track progress:

Task Progress:
- [ ] Step 1: Parse input — extract PR URL, issue URL, commit SHAs
- [ ] Step 2: Collect context — fetch commits, diff, existing descriptions
- [ ] Step 3: Analyze changes — categorize, group, build key files table
- [ ] Step 4: Generate PR description
- [ ] Step 5: Handle issue — create, update, or skip
- [ ] Step 6: Apply changes via gh CLI
- [ ] Step 7: Verify and report results to user

Step 1: Parse Input

Extract from user message:

  • PR URL: https://github.com/owner/repo/pull/123 or #123
  • Issue URL: https://github.com/owner/repo/issues/456 or #456
  • Commit SHAs: full or short SHA hashes

Decision tree:

IF PR URL provided:
  → Continue to Step 2

IF only commit SHAs provided (no PR):
  → ASK user: "Should I also create/update an issue, or just analyze the commits?"
  → If user provides issue link → continue with both
  → If user says skip → analyze commits only and present summary

IF only issue URL provided (no PR):
  → ASK user: "Please provide a PR URL or commit SHAs to analyze changes"

IF nothing provided:
  → ASK user: "Please provide a PR URL, issue URL, or commit SHAs"

Step 2: Collect Context

Gather all necessary data:

For PR:

gh pr view {number} --json title,body,state,baseRefName,headRefName,commits
gh pr diff {number}

For specific commits:

git show {sha} --stat          # files changed
git show {sha} --format=""     # full diff
git log {sha} -1 --format="%H %s%n%b"  # commit message

For existing issue (if provided):

gh issue view {number} --json title,body,state

For parent issue detection:

  • Check PR body for Closes #N, Fixes #N, Parent issue: #N
  • Check issue body for Parent issue: #N
  • If parent issue found, fetch its body too:
gh issue view {parent_number} --json title,body

Step 3: Analyze Changes

Systematically analyze the collected diffs:

3.1. Categorize each change:

  • refactor — restructuring without behavior change
  • feature — new functionality
  • bugfix — fixing broken behavior
  • cleanup — removing dead code, unused imports
  • i18n — translations, localization
  • test — test additions/modifications
  • docs — documentation changes
  • style — formatting, naming, import ordering

3.2. Group by logical area:

  • Identify related changes across files
  • Group into named sections (e.g., "StepStore — Input/Output Consolidation")
  • Each section should have a clear purpose statement

3.3. Identify key patterns:

  • Deleted files (important to highlight)
  • Renamed fields/methods (before → after)
  • New APIs/interfaces
  • Breaking changes
  • Temporary/hardcoded values (mark for follow-up)

3.4. Build Key Files table:

FileChange
path/to/file.tsBrief description
path/to/deleted.tsDeleted

Step 4: Generate PR Description

Use this structure:

## Summary

[2-3 sentences: WHAT was done, WHY it was done. Focus on the purpose, not individual changes.]

Closes #N

## Changes

### [Section Name 1]
- Bullet point describing specific change
- Another bullet point
- Use `code formatting` for identifiers

### [Section Name 2]
...

### Other
- Minor changes that don't warrant their own section

## Key Files

| File | Change |
|------|--------|
| `path/file.ts` | Brief description |

PR description rules:

  • Summary: concise, 2-3 sentences max
  • Changes: grouped by logical area, not by file
  • Use backticks for code identifiers (ClassName, methodName, fileName.ts)
  • Highlight deleted files with Deleted in bold
  • Include Closes #N if issue is linked
  • Key Files table: sorted by importance, not alphabetically

Step 5: Handle Issue

5.1. If issue URL was provided by user:

Fetch existing issue body and compare with PR changes:

IF existing body is empty:
  → Generate full issue body and apply

IF existing body has content:
  → Check for contradictions:
    - Does the existing description claim something different from what the diff shows?
    - Are there sections that describe changes not present in the diff?
    - Are there changes in the diff not covered by existing sections?
  
  IF contradictions found:
    → Present contradictions to user
    → ASK: "I found contradictions in the existing issue description. Should I:
       1. Update the contradicting sections and add missing ones
       2. Replace the entire description
       3. Only append new sections"
    → Apply user's choice

  IF no contradictions (only missing sections):
    → Add new sections to existing body
    → Inform user what was added

5.2. If no issue URL provided but PR is linked to a parent issue:

ASK user: "PR is linked to parent issue #N. Should I:
  1. Create a sub-issue under #N with detailed description
  2. Link to an existing issue (provide number)
  3. Skip issue documentation"

IF user chooses to create sub-issue:
  → Generate issue title from changes
  → Generate detailed issue body (numbered sections, more detail than PR)
  → gh issue create
  → Update parent issue body with new sub-issue link
  → Update PR body with "Closes #NEW_ISSUE"

5.3. If no issue URL and no parent issue detected:

ASK user: "No linked issue found. Should I:
  1. Create a new issue for this PR
  2. Create a sub-issue under an existing parent (provide parent issue number)
  3. Skip issue documentation"

Issue body format (more detailed than PR):

## Summary

[Same as PR summary but expanded with more context]

**PR**: #N
Parent issue: #M (if applicable)

---

## Changes

### 1. [Section Name]

[Detailed paragraph explaining what was changed and why]

- Specific bullet points with technical details
- Before → after descriptions for renames/refactors
- Code examples if helpful

### 2. [Section Name]
...

---

## Key Files

| File | Change |
|------|--------|
| `path/file.ts` | Description |

Step 6: Apply Changes

Execute the changes via gh CLI:

# Update PR description
gh pr edit {number} --body "..."

# Update existing issue
gh issue edit {number} --body "..."

# Create new issue
gh issue create --title "..." --body "..."

# Update parent issue (append sub-issue link)
gh issue edit {parent_number} --body "..."

Important:

  • Use heredoc (cat <<'EOF') for body content to preserve formatting
  • Always verify the result after each operation

Step 7: Verify & Report

After all operations, report to the user:

Done. Here's what was created/updated:

- PR #N description updated: {url}
- Issue #M created/updated: {url}
- Parent issue #P updated with sub-issue link: {url}

Contradiction Detection Rules

When comparing existing issue description with PR diff:

  1. Scope contradiction: Description says change affects only X types, but diff shows it affects all types
  2. Field name contradiction: Description uses old field name, diff shows it was renamed
  3. Missing sections: Diff contains changes not described in any section
  4. Stale sections: Description contains sections about changes not in the diff
  5. Required/optional contradiction: Description says field is optional, diff shows it's required

Always present contradictions to user before making changes.

Quality Standards

  • PR description: concise, scannable, well-structured
  • Issue description: detailed, comprehensive, with numbered sections
  • Always use English for code documentation
  • Never invent changes not present in the diff
  • Always verify diff content before writing descriptions
  • Group related changes logically, not by file
  • Highlight breaking changes and deleted files prominently
  • Mark temporary/hardcoded values for follow-up

Error Handling

ErrorAction
PR not foundAsk user to verify PR number/URL
Issue not foundAsk user to verify issue number/URL
Commit SHA not foundAsk user to verify SHA
gh CLI not authenticatedInstruct user to run gh auth login
Empty diffInform user that PR has no file changes
Permission deniedInform user they may not have write access

Rules of Engagement

  1. DO analyze ALL commits in the PR, not just the latest one
  2. DO check for contradictions before updating existing descriptions
  3. DO ask user before overwriting existing issue content
  4. DO update parent issue when creating sub-issues
  5. DO include Closes #N in PR body when issue is linked
  6. DO NOT invent or assume changes not visible in the diff
  7. DO NOT silently overwrite existing issue descriptions
  8. DO NOT create issues without user confirmation
  9. DO NOT modify PR title unless explicitly asked
  10. DO NOT write comments on GitHub PRs/issues (only edit body)

Job Context Awareness

If called within an orchestrator job context, check for job context before starting:

IF JOB_NAME is provided AND jobs/<JOB_NAME>/ai/context.md exists:
  Read context.md — use it to understand the codebase, affected areas, and conventions
  This enriches PR description generation with project-specific knowledge
ELSE:
  Proceed with standard workflow (analyze PR/commits directly)

The job context path is: <JOBS_ROOT>/<JOB_NAME>/ai/context.md

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.