skills.rest
lapaixkemsdortshlee-svg
lapaixkemsdortshlee-svgCOMMUNITY
Explore all skills by this author →
GitHub

agent-watchdog

Audits another agent's session or work output against the original request.

Updated Apr 3, 2026
Search Tags:#code review#pull request#verification#transcript analysis#agent audit#session monitoring
One-click install
npx skills add https://github.com/lapaixkemsdortshlee-svg/AyitiMarket --skill agent-watchdog
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: agent-watchdog
Source: https://github.com/lapaixkemsdortshlee-svg/AyitiMarket/tree/main/.agents/skills/agent-watchdog
Command: npx skills add https://github.com/lapaixkemsdortshlee-svg/AyitiMarket --skill agent-watchdog

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill helps you verify whether another agent actually completed the work it claimed, instead of relying on summaries or assumptions.

Core Features & Use Cases

  • Watch and audit agent sessions, transcripts, PRs, branches, logs, and run summaries.
  • Compare outcomes across multiple agents when the same request was handled in different ways.
  • Check evidence such as diffs, tests, CI results, screenshots, and review state to spot gaps, scope drift, or weak verification.
  • Targeted repair can be applied when authorization is given and the missing work is narrowly scoped.

Quick Start

Use the agent-watchdog skill to audit this agent session, compare it with the original request, and report any gaps or residual risks.

Frequently Asked Questions about agent-watchdog

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I audit an agent session against the original request?▼

You audit an agent session by resolving the supplied transcript or log artifact, reconstructing the original request contract, and inspecting verification evidence like diffs to report any gaps or scope drift.

What is the best way to verify if an agent's pull request actually matches the requested work?▼

Verifying a pull request involves checking evidence like diffs, tests, and CI results against the reconstructed original request to spot gaps, scope drift, or weak verification in the agent's work.

Can I use this approach to compare outcomes across multiple agents handling the same request?▼

Yes, you can compare outcomes across multiple agents by inspecting their respective verification evidence and diffs to identify gaps, scope drift, or residual risks when the same request was handled differently.

Does this work with Codex and Claude Code transcripts?▼

Yes, this auditing process works with Codex, Claude Code, branches, and run summaries by resolving the supplied transcript artifact to inspect diffs and report gaps, scope drift, or residual risk.

How do I apply targeted repair when an agent audit reveals missing work?▼

You apply targeted repair by providing authorization for narrowly scoped missing work, which allows the process to apply optional narrow fixes to resolve the identified gaps or residual risks.

When do I need to reconstruct a contract during an agent audit?▼

You need to reconstruct a contract when resolving a supplied artifact like a log or transcript to verify the original request, enabling you to inspect diffs and report gaps, scope drift, or residual risk.

Related Skills

Explore More Software Engineering Skills→
cacr92cacr92
4

merge-main-cleanup

Merge the current Git branch into main and prune all other branches.

Community
Advanced
mostafa-ismail-2004mostafa-ismail-2004

terminal-ops

Execute evidence-first repository operations with terminal and git verification.

Community
Advanced
gckogcko
1

story-review-loop-cope

Automate three-agent code review for design/story documentation PRs.

Community
Advanced
FerroxLabsFerroxLabs
21

ferrox-resume-work

Restore project state from STATE.md and resume unfinished workflows.

Community
Intermediate