AgentSkillsCN

logs

显示 Evoloop 管道运行的日志输出——最新运行摘要、特定故事日志、上一次代理调用,或某次具体运行的详细记录

SKILL.md
--- frontmatter
name: logs
description: Show log output from Evoloop pipeline runs — latest run summary, specific story logs, last agent invocation, or a specific run
argument-hint: "[US-XXX | last | run-YYYYMMDD-HHMMSS]"
allowed-tools:
  - Read
  - Glob
  - Grep

Run Log Viewer

Mode (from $ARGUMENTS)

ArgumentMode
(empty)Latest run summary
lastLast agent invocation from most recent run
run-YYYYMMDD-HHMMSSSpecific run log
US-XXXAll log entries for that story across runs
anything elseUsage help

Log Structure

  • .log/run-YYYYMMDD-HHMMSS/run.md — agent invocations
  • .log/run-YYYYMMDD-HHMMSS/context-pack.md — files given to agents
  • .log/run-<id>/handoff-notes.md — may exist per run
  • .state/pipeline.jsonlastRunId

Key patterns in run.md: Runner: <agent> -> <command>, Agent failure: <agent> exit=<code> story=<id>, ## Crashed, Start:, Resume:

Latest Run (no args)

Read pipeline.json for lastRunId, then run.md + context-pack.md. Show: run ID, start time, resume status, agent activity (agent/story/outcome), context pack, pipeline state.

Last Agent (last)

Find latest run, extract last Runner: line + any Agent failure: line. Show: run, agent, runner command, story, outcome.

Specific Run (run-...)

Same as latest run format. If >200 lines → first 30 + "truncated" + last 30. If not found → list available runs.

Story Logs (US-XXX)

Grep all run.md and context-pack.md for story ID. Per matching run: run ID, agent, outcome, stage. Reverse chronological.

Errors

No runs → "Run ./orchestrator.sh run --tool claude". Run not found → list available. No matches → "No entries for {ID}".