Ormus-SolutionsOrmus-Solutions/ormus-agent-ops/skills/agent-inspect/SKILL.md

agent-inspect

Skill0 starsChanged 7 months agoMITView on GitHub

Deep dive into a single task in an autonomous coding agent — details, log, events, dependencies, review verdict. Use when a task's behavior is mysterious or `/agent-triage` couldn't classify it.

2 other projects use a version of this.

Adopted by

Most recently taken up by

rajudandigam/agent-inspecttheir own version, first seen Aug 5, 2026
HermeticOrmus/ormus-agent-opsword for word, first seen Apr 25, 2026

Other versions2 projects use a version of this.
Word for word · 1
HermeticOrmus/ormus-agent-ops★ 0100% same
Edited · 0
None found.
Wrote their own version · 1
rajudandigam/agent-inspect★ 211same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository6 other files it gives its agents.
- agent-cleanSkill
- agent-evolveSkill
- agent-opsSkill
- agent-statusSkill
- agent-sweep-orphansSkill
- agent-triageSkill

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.

≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.

Common
Distinctive
Very specific

Distinctive:
Read the last 100 lines for failures.
line 63
Distinctive:
If your agent runs review on output:
line 67
Distinctive:
If diagnosis is uncertain, say so.
line 153

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.

Files

SKILL.md

/agent-inspect — Deep Task Inspection

When a task is weird, this is where you go.

Prerequisites

Var	Purpose
`AGENT_API_URL`	Base URL of the agent's HTTP API
`AGENT_API_TOKEN`	Bearer token

Endpoints used (adapt to your agent's API):

GET /api/tasks/:id — task metadata
GET /api/tasks/:id/log — execution log
GET /api/tasks/:id/events — event stream
GET /api/tasks/:id/review — review verdict (if your agent runs review)
GET /api/tasks/:id/deps — task dependencies / dependents

Usage

/agent-inspect <task-id>

Workflow

1. Pull metadata

curl -s -H "Authorization: Bearer $AGENT_API_TOKEN" \
  "$AGENT_API_URL/api/tasks/<task-id>" | jq

Capture: status, prompt, model, project, created_at, updated_at, attempt count, parent task (if subtask), priority, cost.

2. Pull events timeline

curl -s -H "Authorization: Bearer $AGENT_API_TOKEN" \
  "$AGENT_API_URL/api/tasks/<task-id>/events" | jq

Build a timeline. Look for:

State transitions (queued → running → review → done/failed)
Retry attempts
Worker reassignments
Review pass/fail rounds
External events (was the agent restarted? did the worker die?)

3. Pull log

curl -s -H "Authorization: Bearer $AGENT_API_TOKEN" \
  "$AGENT_API_URL/api/tasks/<task-id>/log"

Read the last 100 lines for failures. For successful but mysterious tasks, scan for unusual patterns: long pauses, retries, tool errors, low-confidence verdicts.

4. Pull review verdict (if applicable)

If your agent runs review on output:

curl -s -H "Authorization: Bearer $AGENT_API_TOKEN" \
  "$AGENT_API_URL/api/tasks/<task-id>/review"

Capture: verdict (PASS/WARN/FAIL/NEEDS_REVIEW), reviewer model, specific issues raised.

5. Pull dependency graph

curl -s -H "Authorization: Bearer $AGENT_API_TOKEN" \
  "$AGENT_API_URL/api/tasks/<task-id>/deps"

Identify:

Parent task (if this is a subtask)
Sibling subtasks (other subtasks of the same parent)
Dependent tasks (what's waiting on this one)

Output format

TASK INSPECTION — <task-id>
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

METADATA
  status:     <state>
  project:    <project>
  model:      <model>
  attempt:    <N> of <max>
  created:    <relative time>
  updated:    <relative time>
  duration:   <total / current>
  cost:       $<amount>
  parent:     <parent-task-id or "—">

PROMPT
  <first 200 chars of prompt; truncate if longer>

TIMELINE
  <ISO time>  queued
  <ISO time>  running (worker <N>)
  <ISO time>  review-requested
  <ISO time>  review-failed (<reason>)
  <ISO time>  retried
  <ISO time>  failed

REVIEW VERDICT
  verdict:    <PASS/WARN/FAIL/NEEDS_REVIEW>
  reviewer:   <model>
  issues:
    • <issue 1>
    • <issue 2>

DEPENDENCIES
  parent:     <parent or "—">
  siblings:   <count>  (passed: <N>, failed: <N>, pending: <N>)
  dependents: <count>  (waiting on this task)

LOG TAIL (last 30 lines)
  [paste last 30 lines, with line numbers]

DIAGNOSIS
  [1-3 sentences synthesizing what likely happened, based on timeline + log + verdict]

RECOMMENDED ACTION
  [specific next step: requeue / dead-letter / amend prompt / fix env / inspect parent]
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Diagnostic patterns

Symptom	Likely cause	Action
Task stuck in `running` for > expected time	Worker hung or zombie	Stop + requeue
Multiple `retried` events but same failure	Same root cause not fixed	Read log carefully, address cause
Review keeps failing with same complaint	Prompt under-specified	Amend prompt, requeue
Subtask succeeds but parent fails	Aggregation bug or last subtask broken	Inspect last subtask
All siblings passed, this one failed	Task-specific issue, not systemic	Look at this task's prompt vs siblings

Notes

Don't truncate the log until you've read it. A 500-line log with the answer in line 247 means truncation lies.
For very long logs (>2000 lines), summarize structurally (count of each event type) and quote only the last 30 lines verbatim.
If diagnosis is uncertain, say so. Don't fabricate a cause.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

It worked
Partly
It didn't

What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free