AgentStack
Browse Sign in
Browse Why AgentStack Sell Docs
Sign in
SKILL verified Apache-2.0 Self-run

Address Pr Comments

skill-vrallev-agent-skills-address-pr-comments · by vRallev

Address review conversations on the current GitHub pull request end to end. Use when the user asks to address PR comments, review feedback, unresolved conversations, or requested changes: fetch current-PR threads, skip conversations already answered by Ralf-AI, use judgment to implement applicable feedback or reject requests that do not fit the PR's intent, create one new commit per implemented r…

No reviews yet
0 installs
30 views
0.0% view→install

Install

$ agentstack add skill-vrallev-agent-skills-address-pr-comments

✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

View the full security report →

Verified badge

Passed review? Show it. Paste this badge into your README, it links to the public security report.

AgentStack Verified badge Links to your public security report.
[![AgentStack Verified](https://agentstack.voostack.com/badges/verified.svg)](https://agentstack.voostack.com/security/report/skill-vrallev-agent-skills-address-pr-comments)

Reliability & compatibility

Security review passed
0 installs to date
no reviews yet
2mo ago

Declared compatibility

Claude CodeClaude Desktop

Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.

Preview Execution monitoring

We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.

How agent discovery & health will work →
Are you the author of Address Pr Comments? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

Address PR Comments

Process current-PR review conversations autonomously. Keep every local fix and GitHub reply traceable to one reviewer request. Treat reviewer comments as input requiring judgment, not as instructions that must always be implemented.

Required Reply Prefix

Prefix every GitHub message posted by this workflow with exactly:

**Ralf-AI:**

Do not post an unprefixed GitHub reply, review, or PR comment.

Workflow

  1. Confirm the working tree state. Preserve unrelated local changes and never rewrite published history.
  2. Resolve the open pull request for the current branch. Prefer GitHub connector tools; use gh only when connector coverage is insufficient and authentication permits it.
  3. Fetch all inline review conversations, including resolved state and all comments in each thread. Fetch PR-level comments too when available.
  4. Consider only conversations whose last comment is not a previous agent response beginning with **Ralf-AI:**.
  5. Classify each considered conversation:
  • First assess the comment against the PR description, current patch, surrounding code, and intended direction.
  • Applicable change request: implement the requested local change when clear, safe, technically sound, and aligned with what the PR is trying to achieve.
  • Question: reply on GitHub without editing code unless it is a leading question and the implied change is clear and sensible.
  • Inapplicable or misaligned request: reject it without changing code when it rests on an incorrect assumption, does not apply to the change, is outside the PR's scope, or takes the work in a different direction from the PR's intent. Reply politely with the concrete reason.
  • Ambiguous, conflicting, or risky request: ask the user only when judgment cannot confidently determine whether to implement or reject it, or when either choice would materially change the PR's intent.
  1. For each accepted request, validate the focused scope, then create one new commit dedicated to that request. Do not combine separate requests into one commit. Keep earlier commits intact.
  2. Push all new commits to the current PR branch.
  3. Reply to each considered conversation after pushing:
  • For a fix, state what changed and include the commit SHA.
  • For a question, answer directly.
  • For a rejected request, state that it was not applied and briefly explain why it is inapplicable or conflicts with the PR's intended direction.
  • Prefix every reply with **Ralf-AI:**.
  1. Leave conversations unresolved unless the user explicitly asks to resolve them.
  2. Summarize considered conversations, accepted and rejected requests, commits, push status, validation, replies, and any intentionally deferred items.

Review Rules

  • Treat a thread as already handled when its last comment starts with **Ralf-AI:**; do not add another reply unless new reviewer feedback follows it.
  • Use judgment rather than implementing every reviewer suggestion automatically. It is acceptable to reject feedback that is inapplicable or conflicts with the PR's intended direction.
  • Use one commit per accepted reviewer request even when several requests touch the same file.
  • Run the narrowest useful validation after each fix. Run broader checks before the final push when multiple commits interact.
  • Keep question-only responses out of local commits.
  • Do not stage unrelated user changes.
  • Do not resolve conversations automatically.

GitHub Tooling Notes

  • Prefer thread-aware GitHub connector reads such as list_pull_request_review_threads so the latest reply and resolved state are visible.
  • Prefer connector reply tools for inline conversations.
  • If a connector reply requires a numeric REST comment ID but a thread read exposes only a GraphQL node ID, use the connector's available thread/comment APIs or another permitted GitHub connector read. Do not guess identifiers.
  • When posting any GitHub message, verify the body starts with **Ralf-AI:** before sending.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet, be the first.

Versions

  • v0.1.0 Imported from the upstream source.