Pr Context Gatherer

PR Context Gatherer

Use this skill to collect pull request context once and reuse it across review runs.

Inputs

• Pull request number, URL, or branch name (optional).
• If no input is provided, detect the PR from the current branch.

Output

Create .restflow/pr-context/{pr_number}/ with the following files:

1. pr-meta.json

• Command:

gh pr view {PR} --json number,title,body,author,baseRefName,headRefName,state,labels,reviewDecision,mergeable,additions,deletions,changedFiles

2. comments.json

• Command:

gh api repos/{owner}/{repo}/pulls/{PR}/comments --paginate

3. review-threads.json

• Command:

gh api graphql -f query='\
query($owner: String!, $repo: String!, $pr: Int!) {\
  repository(owner: $owner, name: $repo) {\
    pullRequest(number: $pr) {\
      reviewThreads(first: 100) {\
        nodes {\
          id\
          isResolved\
          isOutdated\
          path\
          line\
          startLine\
          comments(first: 20) {\
            nodes {\
              author { login }\
              body\
              createdAt\
            }\
          }\
        }\
      }\
    }\
  }\
}'

4. changed-files.txt

• Command:

gh pr diff {PR} --name-only

5. diffs/{sanitized_filename}.diff

• For each changed file, collect a focused diff:

gh pr diff {PR} -- {file} > diffs/{sanitized_filename}.diff

• Optionally include function-level context for code files:

git diff -W origin/{base_branch}...HEAD -- {file}

6. related-issues.json

• Parse PR body references like #123, fixes #456, closes #789.
• Fetch each related issue with:

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

7. context/

• Copy project conventions when available:
  • CLAUDE.md
  • AGENTS.md
  • nearest directory-level AGENTS.md for changed files

8. CONTEXT_SUMMARY.md

• Include:
  • PR number and title
  • changed file count and list
  • resolved/total review-thread counts
  • inventory of generated context files

Implementation Procedure

1. Resolve PR identity.

• If input contains a URL, extract number.
• If input contains a plain integer, use it directly.
• Otherwise run:

gh pr view --json number

2. Create target directory.

mkdir -p .restflow/pr-context/{PR}/diffs .restflow/pr-context/{PR}/context

3. Gather metadata, comments, review threads, and changed files.

• Save each command output to the corresponding JSON or text file.

4. Generate per-file diffs.

• Iterate through changed-files.txt.
• Sanitize file path separators when writing diff filenames.

5. Collect related issues.

• Extract issue numbers from PR body with regex.
• Deduplicate IDs before fetching.

6. Copy context docs.

• Copy only files that exist.
• Do not fail the workflow if optional docs are missing.

7. Write CONTEXT_SUMMARY.md.

• Record what was fetched and what was unavailable.

Hook Integration

Use this hook pattern to auto-run context gathering for review tasks.

{
  "operation": "create",
  "name": "pr-context-pre-gather",
  "event": "task_started",
  "filter": "review*",
  "action": "run_task",
  "agent": "pr-context-gatherer-agent",
  "input": "Gather context for PR {{task_name}}"
}

CLI equivalent:

restflow hook create \
  --name pr-context-pre-gather \
  --event task_started \
  --action run_task \
  --agent pr-context-gatherer-agent \
  --input "Gather context for PR {{task_name}}"

Consumer Guidance

Review agents should:

1. Check for .restflow/pr-context/{PR}/CONTEXT_SUMMARY.md first.
2. Load all context files in one batch using the file tool.
3. Fall back to direct gh calls only if required files are missing.

Success Criteria

• Context directory exists with expected files.
• A downstream review task can complete initial analysis without repeated metadata fetch calls.
• Missing optional files are reported in summary but do not fail the run.

Converted and distributed by TomeVault — claim your Tome and manage your conversions.