pr-automation-workflows

PR Automation Workflows

CRITICAL: There are FOUR distinct workflows, each running on separate cron schedules. They are NOT conditional branches - they are separate execution modes.

1. Comment-Only Mode (Default)

Cron: Every hour (0 * * * *) Command: jleechanorg-pr-monitor --max-prs 10

What it does:

• Posts "Codex support" comments on ALL non-draft PRs (regardless of branch name)
• Comment format: Asks Codex to provide coding assistance
• Target: Up to 10 PRs per run
• Marker: <!-- codex-automation-commit:SHA -->

Scope: ALL open, non-draft PRs in the organization

2. Comment-Validation Mode

Cron: Every 30 minutes (*/30 * * * *) Command: jleechanorg-pr-monitor --comment-validation --max-prs 10

What it does:

• Posts comment validation requests on ALL non-draft PRs (regardless of branch name)
• Comment format: @coderabbit-ai @greptileai @bugbot @copilot - asks bots to review
• Target: Up to 10 PRs per run
• Marker: <!-- comment-validation-request:SHA -->

Scope: ALL open, non-draft PRs in the organization

3. Fixpr Mode

Cron: Every 30 minutes (*/30 * * * *) Command: jleechanorg-pr-monitor --fixpr --max-prs 10 --cli-agent gemini,cursor

What it does:

• Fixes merge conflicts and failing CI checks
• Uses orchestrated PR runner with Gemini/Cursor agents
• Target: Up to 10 PRs with conflicts or failing checks
• Only processes PRs that have actual issues (conflicts/failing checks)

Scope: PRs with merge conflicts OR failing CI checks

4. Fix-Comment Mode (Currently Disabled)

Cron: Commented out Command: jleechanorg-pr-monitor --fix-comment --cli-agent gemini,cursor --max-prs 3

What it does:

• Resolves PR review comments
• Uses Gemini/Cursor to address bot feedback
• Target: Up to 3 PRs per run

Scope: PRs with unaddressed review comments

Additional Workflows

Codex Update Mode

Cron: Every hour at :15 (15 * * * *) Command: jleechanorg-pr-monitor --codex-update What: Runs Codex automation tasks (browser-based)

Codex API Mode

Cron: Every hour at :30 (30 * * * *) Command: jleechanorg-pr-monitor --codex-api --codex-apply-and-push What: Runs Codex automation via API (not browser)

Key Distinctions

Workflow	Posts Comments?	Branch Filter?	Target PRs
Comment-only	Yes (Codex support)	NO - ALL PRs	All non-draft
Comment-validation	Yes (bot review)	NO - ALL PRs	All non-draft
Fixpr	No (fixes issues)	NO	PRs with conflicts/failing checks
Fix-comment	No (addresses comments)	NO	PRs with review comments

Common Mistakes

❌ WRONG: "Comment-only mode should only post on Codex branches" ✅ RIGHT: "Comment-only mode posts on ALL non-draft PRs"

❌ WRONG: "Comment-validation is for non-Codex branches only" ✅ RIGHT: "Comment-validation is a separate cron job that runs on ALL non-draft PRs"

❌ WRONG: "Use branch name to decide which comment type to post" ✅ RIGHT: "The cron job determines the comment type (--comment-validation flag)"

Implementation

The monitoring cycle code path (line 4003 in jleechanorg_pr_monitor.py):

if comment_validation:
    # Post comment validation request (asks AI bots to review)
    comment_result = self.post_comment_validation_request(repo_full_name, pr_number, pr)
else:
    # Post codex instruction comment (asks Codex to code)
    comment_result = self.post_codex_instruction_simple(repo_full_name, pr_number, pr)

The flag determines the comment type, NOT the branch name.

Copilot and Codex Tracking Requirements

When automation (Codex) or the /copilot tool fixes PR comments, require URL tracking artifacts:

• responses.json entries should include html_url for each processed comment.
• Tracking commits should include the appropriate marker:
  • [codex-automation-commit] for automation system tasks.
  • [copilot-commit] for /copilot tool executions.
• Commits should use separate URL buckets for:
  • FIXED
  • CONSIDERED (ACKNOWLEDGED/DEFERRED/NOT_DONE)
• If responses.json is missing URLs, backfill from comments.json by matching comment IDs in a type-safe way ((.id|tostring) == $id).

This keeps both fix-comment/codex automation and manual /copilot workflows auditable and consistent with expectations.