CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

agent-watchdog

Use when asked to watch, babysit, audit, review, compare, or fix another agent's work from a Codex session ID, Claude Code session/transcript, chat/thread link, PR, branch, log, or pasted run summary. Monitor until the other agent is done or blocked, reconstruct what the user asked, independently investigate the same problem to form your own hypotheses and approach, inspect what the agent actually changed and verified, compare the two investigations, report gaps and add-on directions, and optionally make scoped fixes when the user authorizes repair.

Quality

92%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

SKILL.md
Quality
Evals
Security

Quality

Content

85%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body is a tightly written, well-sequenced process guide with concrete classification schemes, explicit validation checkpoints, and a ready-to-use report template; it wastes almost no tokens on things Claude already knows. Remaining gains are minor: trim the motivational framing and add a short example of the batched relay note it describes.

Suggestions

Trim the figurative framing sentences ('reviewer with a pager', 'not just grading their homework') to pure instruction — the role definition lands in two lines instead of six.

Add a two- or three-line example of the batched/ranked relay note format, since 'Relay Sparingly' is the most novel section and the only one with no concrete artifact.

Optionally include a one-line example of an 'add-on direction' (file + guard + test case) to make the diff-output requirement concrete.

DimensionReasoningScore

Conciseness

The body is dense with non-obvious process guidance (anchoring failure mode, relay economics, issue taxonomy) and explains nothing Claude already knows, but a few rhetorical flourishes ('like a reviewer with a pager', 'You are not just grading their homework') are motivational padding that could be trimmed. This is the score-4 anchor ('efficient; minor instances of over-explanation'), not 5.

4 / 5

Actionability

Concrete, executable instruction throughout: numbered procedures for each phase, a five-way issue classification, a copy-paste-ready report template, and specific directives ('reopen the cited files and line refs', 'Re-run the smallest useful validation after each meaningful fix'). It falls short of the score-5 anchor only because formats like the 'batched relay note' and 'add-on directions' are described but never exemplified.

4 / 5

Workflow Clarity

A clear mode-selection → target-resolution → contract → investigation → audit → fix → report sequence with explicit validation checkpoints ('Verify the watched agent's key claims against primary sources', 'Re-run the smallest useful validation after each meaningful fix') and error-recovery stop conditions ('If a fix would require a product decision... stop and report'). This matches the score-5 anchor with feedback loops and checklists.

5 / 5

Progressive Disclosure

No bundle files exist (no references/, scripts/, or assets/ directories), and nothing in the body needs splitting out — every section is process guidance that belongs in SKILL.md. Section headers are clear and well-ordered, references are non-nested, and navigation is trivial, fitting the clean-structure score-5 anchor.

5 / 5

Total

18

/

20

Passed

Description

96%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong description: it explicitly pairs a rich 'Use when' trigger clause with a comprehensive, concrete list of capabilities, and names the specific artifact types (session IDs, transcripts, thread links, PRs) that identify the skill. Its only weakness is mild overlap risk with generic review/PR-audit skills from words like 'audit', 'review', and 'fix'.

Suggestions

Sharpen distinctiveness by leading the trigger list with the most distinctive phrasing (e.g., 'Use when asked to oversee or second-check another agent's (Codex or Claude Code) work') so generic 'review my PR' requests are less likely to match.

Consider adding one or two everyday synonyms users might say ('supervise', 'keep an eye on', 'second opinion on') to the trigger list.

DimensionReasoningScore

Specificity

The description enumerates seven-plus concrete actions ('Monitor until the other agent is done or blocked, reconstruct what the user asked, independently investigate the same problem..., inspect what the agent actually changed and verified, compare the two investigations, report gaps..., and optionally make scoped fixes'), which matches the comprehensive-coverage anchor; it is not the score-4 anchor because no meaningful capability is missing.

5 / 5

Completeness

It opens with an explicit 'Use when asked to...' clause naming concrete trigger scenarios and follows with a full 'what' list ending in 'when the user authorizes repair'. Both what and when are explicit with concrete trigger phrases — the score-5 anchor; the score-4 anchor ('when could be more explicit') does not apply since the triggers are enumerated.

5 / 5

Trigger Term Quality

Natural trigger verbs ('watch, babysit, audit, review, compare, or fix') are covered with synonyms, and the artifact list ('Codex session ID, Claude Code session/transcript, chat/thread link, PR, branch, log, or pasted run summary') covers the concrete forms users would name. Comprehensive synonym coverage fits the score-5 anchor.

5 / 5

Distinctiveness Conflict Risk

The niche (watchdogging another agent's resolved session/transcript/PR) is distinct, but generic terms like 'audit, review, compare, or fix... PR, branch' create minor overlap with code-review and PR-review skills. This fits 'mostly distinct; minor overlap risk with closely related skills' (4) better than the minimal-conflict score-5 anchor.

4 / 5

Total

19

/

20

Passed

Validation

100%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository

BuilderIO/skills

Path

skills/agent-watchdog/SKILL.md

Reviewed

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.