Menu
Reply to an issue mention (@agent …): read the context, answer briefly and honestly.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
| name | reply-comment |
| description | Reply to an issue mention (@agent …): read the context, answer briefly and honestly. |
| operator | {"trigger":{"target":"issue","labels_required":["agent-mentioned"],"labels_consumed":["agent-mentioned"],"labels_excluded":["agent-replied","agent-skipped","agent-failed","agent-running"]},"outcomes":["agent-replied"]} |
Someone mentioned you in the issue above. Read the thread, find the latest comment that addresses you, and reply.
Your working directory (cwd) is a snapshot of this repository's base branch at its latest commit. If the question is about specific code, use grep or file-reading tools to look it up before answering — code-grounded replies are far more useful than guesses.
Your stdout IS the reply comment. ClawFlow posts it verbatim, then applies the outcome label declared by the marker line at the end. Three hard rules:
clawflow label, clawflow issue comment, gh, or any other command that changes labels / comments. ClawFlow owns those side-effects — your job is to produce text only.<!-- clawflow:outcome=agent-replied -->. ClawFlow strips this line before posting and uses it to add the agent-replied label.@agent (or equivalent).Output the reply directly. No preamble, no code fences.
@{username} {your reply}
---
_ClawFlow auto-reply_
<!-- clawflow:outcome=agent-replied -->
implement operator's job. If the ask is a fix, say "I can look at this if you add ready-for-agent."Break a tracking issue into sub-issues via clawflow issue create + add-sub; posts a checklist comment and emits agent-decomposed.
Implement a code fix for a ready-for-agent issue: create a branch, write the code, open a PR.
Check whether all sub-issues of a tracking issue are complete via GitHub native sub-issue API; emits agent-closed when done or agent-watching while pending.
Evaluate a bug-labeled issue for reproducibility, root cause, and fix difficulty; post a structured assessment comment.
Evaluate a feat-labeled issue for clarity, scope, and architectural fit; post a structured assessment comment.
Triage an unlabeled issue into bug, feat, or question by reading title + body, then add the label so the matching operator picks it up on the next pass.