swarm-pr-feedback

Ingest, verify, and resolve pull request feedback with a tracked closure ledger.

Updated May 31, 2026
One-click install
npx skills add https://github.com/AlexanderNarbaev/agi --skill swarm-pr-feedback-alexandernarbaev
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: swarm-pr-feedback
Source: https://github.com/AlexanderNarbaev/agi/tree/main/.opencode/skills/swarm-pr-feedback
Command: npx skills add https://github.com/AlexanderNarbaev/agi --skill swarm-pr-feedback-alexandernarbaev

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Pull request feedback arrives scattered across review comments, bot reviews, CI failures, merge conflicts, and pasted notes, and items get silently dropped or re-triaged from scratch on every bot review round. This Skill ingests every feedback surface into a single ledger, verifies each claim against the actual code, fixes confirmed issues, and reports closure status for every item. ## Core Features & Use Cases - Feedback Ledger Triage: Normalizes every comment, CI failure, and conflict into FB-### items with statuses like CONFIRMED, DISPROVED, PRE_EXISTING, and NEEDS_USER_DECISION so nothing is lost. - Multi-Round Bot Review Handling: Continues the same ledger across repeated bot review rounds, carries forward unresolved items, and applies a 3-strikes rule for repeatedly raised findings. - Skeptical Source Verification: Checks bot claims about imports, line numbers, ordering, and DI seams against the actual branch before editing, avoiding wrong fixes. - Use Case: A bot reviewer posts new findings after every push to your PR. Use this Skill to triage each round incrementally, fix confirmed issues, validate the branch, and post a per-round closure ledger as a PR comment. ## Quick Start Resolve all known feedback on the current pull request and post a closure ledger summarizing every item's outcome.

Frequently Asked Questions about swarm-pr-feedback

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I address all PR review comments without missing any?

Build a feedback ledger that assigns every review comment, CI failure, and conflict an FB-### ID before editing. Each item gets a verified status, and the final closure ledger must account for every entry so nothing ships unreviewed.

How to handle bot reviewers that re-review after every push?

Continue the same feedback ledger across rounds instead of restarting triage. Carry forward unresolved items, mint fresh IDs only for new findings, and apply a 3-strikes rule that adds defense-in-depth fixes for repeatedly raised findings.

Should I trust bot review claims about my code?

No, treat every bot claim as unverified until checked against the current branch. Bots misread import paths, cite stale line numbers, and miss parent-block guards, so verify each claim with source evidence before editing.

What is the difference between PR review and PR feedback resolution?

PR review discovers new findings through broad inspection, while feedback resolution ingests existing comments and closes them. This Skill handles only the feedback-closure side and explicitly avoids running a fresh broad review.

When should I not apply a bot-suggested fix?

Skip the fix when it would add incorrect, misleading, or redundant code, such as a guard that already exists at an outer scope. Document the disproof with evidence and surface persistent disagreement to the user instead of pushing wrong changes.