CtrlK
BlogDocsLog inGet started
Tessl Logo

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.

72

Quality

87%

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

75%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.

A well-crafted, dense instruction skill with a clear multi-step workflow, concrete issue taxonomy, report template, and validation gates for the destructive fix path. It assumes Claude's competence and avoids concept padding; the main gap is a lack of explicit validate→fix→retry loops and any external reference structure that would push it to the top anchor.

Suggestions

Add explicit validate→fix→retry loops in 'Fix Narrowly' (e.g., 'If validation fails, fix and re-validate; only proceed when valid') to reach the top workflow_clarity anchor.

Tighten the prose in 'Investigate Independently' and 'Relay Sparingly' by converting a few explanatory sentences into terse bullets.

Consider extracting the report template and issue taxonomy into a one-level-deep reference file (e.g., REPORT_TEMPLATE.md) to strengthen progressive disclosure navigation.

DimensionReasoningScore

Conciseness

Lean and purposeful with no padding over concepts Claude already knows (e.g., it never explains what a PR or CI run is), though a few prose passages in 'Investigate Independently' and 'Relay Sparingly' could be tightened slightly. Not a 5 because minor elaborative sentences remain.

4 / 5

Actionability

Concrete, specific guidance throughout — a defined issue taxonomy (Gap/Bug/Verification miss/Scope drift/No issue), a copy-ready report template, and explicit fix constraints — though a few directives like 'Explore the code, data, logs, production state' stay somewhat abstract. Instruction-only, so absence of code is not penalized.

4 / 5

Workflow Clarity

A clear sequenced workflow (Choose Mode → Resolve Target → Reconstruct Contract → Investigate Independently → Audit Evidence → Fix Narrowly → Report) with validation checkpoints ('Re-run the smallest useful validation after each meaningful fix', 'stop and report' for destructive actions). Not a 5 because some checkpoints are implicit rather than enumerated as explicit validate→fix→retry loops.

4 / 5

Progressive Disclosure

A single self-contained file with clear, well-labeled section headers and no nested references; no bundle files exist to verify. Good organization, though the embedded report template and classification scheme could justify a separate reference file. Not a 5 because there is no explicit one-level-deep reference structure to signal navigation.

4 / 5

Total

16

/

20

Passed

Description

100%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.

An exemplary description: concrete actions, natural trigger terms with synonyms and source formats, explicit 'Use when' guidance, and a distinct niche. Voice is imperative/third-person, matching the good examples. No weak dimensions.

DimensionReasoningScore

Specificity

Lists multiple concrete actions — "watch, babysit, audit, review, compare, or fix", "reconstruct what the user asked", "independently investigate", "inspect what the agent actually changed and verified", "compare the two investigations", "report gaps and add-on directions", "make scoped fixes" — with comprehensive coverage of the watchdog task.

5 / 5

Completeness

Explicitly answers both: "Use when asked to..." gives the when with concrete trigger phrases, and the rest of the sentence lays out the full what (monitor, reconstruct, investigate, inspect, compare, report, fix).

5 / 5

Trigger Term Quality

Comprehensive natural triggers ("watch, babysit, audit, review, compare, or fix") plus concrete source synonyms ("Codex session ID, Claude Code session/transcript, chat/thread link, PR, branch, log, or pasted run summary") that a user would naturally say.

5 / 5

Distinctiveness Conflict Risk

Occupies a clear niche — second-pass auditing/investigation of another agent's work — with distinct triggers tied to session IDs, transcripts, PRs, and run summaries, minimizing overlap with generic review or coding skills.

5 / 5

Total

20

/

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.

Validation16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository
BuilderIO/skills
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.