Loop Triage Skill
Generate concise, actionable triage reports for AI coding loops, focusing on high-priority items.
The skill is primarily instructional with no external operations, so risk of permission abuse is low. However, it lacks explicit guidance on sensitive data handling, least privilege, user confirmation, and rollback, hence deduction.
Static review without execution limits the score; the skill has no built-in error handling or tests.
Inputs, outputs, and triggers are clear, but boundaries are not explicit, and no adaptation for Chinese users or mainland China networks.
Repository has MIT license and well-structured docs, but the skill itself lacks version info, changelog, and maintenance ownership.
Effectiveness cannot be verified statically; output format lacks concrete examples.
Evidence from README and CI, but no direct third-party execution evidence of this specific skill's functionality.
- The skill depends on external services (CI, Slack, Linear) that may be unreachable from mainland China, affecting usability.
- No explicit guidance on handling sensitive data; users should be cautious with credentials.
What does this skill do, and when should you use it?
Loop Triage is a skill from the loop-engineering repository that triages recent changes, CI failures, issues, and conversations. It produces a concise, prioritized Markdown report that a loop can consume, and writes structured output to a state file or Linear board. The skill follows strict rules: it only surfaces items that a reasonable engineer would want to know about today, places everything else in watch or ignore, and never proposes architectural overhauls. It is designed for any Agent Skills client, located at skills/loop-triage/SKILL.md.
The skill takes inputs provided by the loop: recent CI/test failures (last 24h), open issues/Linear tickets assigned to the team, recent commits on main (last 24–48h), Slack/chat threads the loop has visibility into, and the current state file. It analyzes these and generates a Markdown report with four sections: High-Priority Items (with impact, suggested action, and rough effort), Watch Items, Noise/Ignore, and State Updates (facts for the loop to remember for next run). Output is concise and actionable, written to a state file or Linear board.
- A developer wants a daily prioritized list of changes, CI failures, and issues for a loop to act on.
- A maintainer runs a daily loop that automatically identifies PRs or issues needing human attention instead of manual review.
- A team wants to track what the loop already knows via state updates and new signals in a machine-readable format.
- A project lead needs a quick overview of what requires immediate action without digging into details.
- Integrating triage as part of a larger loop workflow, such as the daily-triage pattern, running in L1 report mode.
What are this skill's strengths and limitations?
- Provides structured, concise reports that loops can easily consume.
- Clear prioritization rules reduce noise and avoid creating unnecessary work.
- Includes state updates to maintain memory and context across loop runs.
- Tool-agnostic and applicable to a wide range of AI coding agents.
- Part of a broader loop engineering methodology with complementary patterns.
- Depends on external inputs (CI failures, issues, chat threads) that must be provided by the loop or user.
- No built-in test suite; output quality depends on the quality of provided inputs.
- Compatibility with untested platforms like ChatGPT is unknown; verified on Claude Code and Codex.
- Does not perform automatic fixes or actions; it only reports and suggests.
How do you install this skill?
This skill is part of the loop-engineering repository, which is MIT-licensed. The skill file is at skills/loop-triage/. To install, clone the repository or copy the skills/loop-triage/ folder into your project's skills directory. There is no standalone package manager installation command provided.
How do you use this skill?
In an Agent Skills client (e.g., Claude Code, Codex), place the skill folder in your project's skills directory. Then ask the loop to run it, providing the expected inputs as part of the loop context. Example trigger prompt:
/loop 1d Run loop-triage. Update STATE.md. No auto-fix in week one.
The skill will produce a Markdown report and write to a state file or Linear board as configured. Refer to the SKILL.md for output format details.