loop-triage

name	loop-triage
description	Use for a Codex daily triage loop that safely collects and classifies CI, issue, PR, commit, and chat signals into report-only handoff state with bounded scope, no auto-fix, and loop-run-log updates.

Daily Triage Loop

Use this skill for repeatable L1 signal collection and safe report-only handoff. Pair it with skills/loop-engineering/SKILL.md for the global loop guardrails.

Inputs

Collect only the signals needed for the requested cadence:

• Recent CI or test failures from the last 24 hours.
• Open team issues, PRs, and review threads.
• Recent main-branch commits from the last 24 to 48 hours.
• Current STATE.md and loop-run-log.md.
• Existing loop budget and pause flags when present.

Output

Append a concise update to STATE.md:

## High Priority
- Item:
  Why now:
  Suggested next action:
  Effort estimate:

## Watch List

## Noise / Ignore

## State Updates

Append one summary line to loop-run-log.md with timestamp, result, evidence, and next action.

Rules

• Keep the loop L1 report-only unless the user explicitly raises it to L2.
• Put only high-signal items in High Priority.
• Do not make architecture changes.
• Do not edit product code.
• If no actionable items exist, record that and exit quickly.
• If a low-risk single-file bugfix candidate is found, create an L2 handoff item; do not auto-fix in week one.
• Escalate medium or high risk, unclear ownership, repeated failures, or denylisted paths to a human.

Prompt Template

Use this shape when dispatching the loop:

Run loop triage. Update STATE.md with High Priority, Watch List, Noise, and
State Updates. Update run timestamp and append loop-run-log.md. If action is
clear, low-risk, and a single-file bugfix candidate, escalate to minimal-fix
only in L2.

Completion Gate

Mark the run complete only when STATE.md and loop-run-log.md are updated, the evidence source is named, and any human-required item is escalated clearly.