| name | code-first-responder |
| description | Guides the agent through a short, repeatable debug loop: capture the failure, isolate the scope, draft a patch plan, and log the decision so the next agent can continue without rereading the entire repo. |
| allowed-tools | run_command read_file search_code write_file |
| metadata | {"legacy_slug":"code_first_responder","legacy_summary":"Deterministic triage flow for failing tests or regressions.","legacy_version":"0.1.0","legacy_author":"ff-terminal-core","legacy_priority":"default","legacy_tags":"[\"code\",\"debugging\",\"testing\"]","legacy_triggers":"[\"bug\",\"failing test\",\"stack trace\",\"regression\",\"error\",\"traceback\"]","legacy_assets":"[\"templates/debug_report_template.md\",\"templates/fix_checklist.md\"]","legacy_recommended_tools":"[\"run_command\",\"read_file\",\"search_code\",\"write_file\"]"} |
Code First Responder
When to use this skill
- Use when the user request matches this skill's domain and capabilities.
- Use when this workflow or toolchain is explicitly requested.
When not to use this skill
- Do not use when another skill is a better direct match for the task.
- Do not use when the request is outside this skill's scope.
When to load this skill
- Incoming task references failing tests, stack traces, or recent merges.
- User explicitly wants a "quick triage", "first look", or "rough fix plan".
- Another skill (e.g., Rapid Research) flagged new code issues that need confirmation.
Guardrails
- Never edit files before the failure is reproduced locally.
- Capture the exact command + exit code inside the debug report template.
- Prefer minimal diffs; defer major refactors to a follow-up task.
- If the failure cannot be reproduced quickly, escalate with the captured evidence and leave the repo untouched.
Five-step micro-playbook
- Snapshot context
- Record branch, latest commit hash, and dirty files.
- Paste the failure summary into
templates/debug_report_template.md.
- Reproduce fast
- Run the narrowest test or script possible.
- Store the exact CLI command and snippet of output.
- Isolate scope
- Identify the component or file set causing the issue.
- Outline suspected root causes without editing yet.
- Propose fix plan
- List concrete actions (edit file X, update test Y, run command Z).
- Mark each action as
must, optional, or blocked.
- Optional patch
- If the fix is obvious (<30 lines), apply it and run validation.
- Update the debug report with the diff summary and test status.
Output checklist
Refer to templates/fix_checklist.md and ensure every box is addressed:
- Failure reproduced?
- Minimal scope understood?
- Plan or patch documented?
- Validation command + result captured?
- Handover note written?
Response snippet to send back to the user
## Debug Status
- Branch/commit: ...
- Failure reproduced: yes/no (command)
- Root-cause hypothesis: ...
- Next actions: ...
## Attachments
- debug_report_template.md (updated)
Companion tools
run_command for fast tests.
read_file + search_code for targeted inspection.
write_file for updating the templates and final report.