CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

autopilot-pr-review-worker

Use this skill to run a multi-persona expert advisory review on a labelled pull request in microsoft/apm. panel-mode (full | lean | delta) selects a surface-gated roster plus a CEO synthesizer; omitted or unknown mode is lean. Do not spawn inactive stubs. The orchestrator is the sole writer to the PR: ONE recommendation comment, no verdict labels, no merge gating. The panel is advisory -- it surfaces findings, prioritizes follow-ups, and renders a ship-recommendation that the maintainer and author weigh.

58

Quality

73%

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

Fix and improve this skill with Tessl

tessl review fix ./packages/autopilot/autopilot-pr-review-worker/SKILL.md

The canonical home for this skill is autopilot-pr-review-worker in microsoft/apm

SKILL.md
Quality
Evals
Security

Quality

Content

81%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.

An exceptionally actionable, well-gated operations manual: every step has exact commands, bounded failure recovery, and explicit validation, and all asset references resolve one level deep. Its weaknesses are redundancy -- the fan-out and non-empty-exit invariants are each stated three times -- and a fast-path/gotcha bulk that belongs in the already-existing assets files.

Suggestions

Consolidate the thrice-stated synchronous-fan-out / non-empty-exit rule into one invariant section and reference it from step 5 and the Gotchas instead of restating it in full.

Move the per-persona fast-path file lists out of the body entirely and keep them only in assets/panel-mode.md, which already duplicates them -- the body should state the gating principle and point to the asset.

Consider folding the mermaid classDiagram gotcha and other single-persona details into the corresponding .agent.md files so the orchestrator body stays at overview altitude.

DimensionReasoningScore

Conciseness

The body is dense and almost entirely load-bearing -- no basic-concept explanations and no filler -- but the same critical rules are restated in full: the synchronous-fan-out / non-empty-exit rule appears three times (architecture invariants, step 5, Gotchas) and "defaults to lean" appears five times. Mostly efficient but could be tightened by consolidating the repeated invariants, which keeps it at anchor 3 rather than 4.

3 / 5

Actionability

Guidance is fully executable throughout: exact commands ("gh pr view <PR_NUMBER> --json author,reviewRequests,headRefOid"), an exact label sweep list ([panel-review, panel-approved, panel-rejected]), an exact receipt line format, verbatim severity-contract text for panelist prompts, byte budgets, truncation limits, and bounded re-spawn counts. The common cases are covered copy-paste ready; the only unspecified call (autopilot-comment) is environment-provided tooling.

5 / 5

Workflow Clarity

The execution checklist (steps 0-9) sequences strictly ("Do not skip ahead. Do not emit any output to the PR before step 6") with validation at every risky point: acceptance gate, CODEOWNERS last-comment gate, schema gates with re-spawn feedback loops (max two attempts, then a placeholder), an ownership consistency gate with re-render-once then fail-closed noop, and an explicit non-empty-exit check. This matches the validate -> fix -> retry pattern of the anchor-5 example.

5 / 5

Progressive Disclosure

All five asset references (panel-mode.md, panelist-return-schema.json, ceo-return-schema.json, recommendation-template.md, shared-brief.example.json) exist in the bundle, are clearly signaled at point of use, and are one level deep. However the body is a ~740-line operations manual rather than an overview, and it inlines the persona fast-path lists that assets/panel-mode.md already carries -- good structure with minor organization gaps, matching anchor 4 rather than the appropriately-split ideal of anchor 5.

4 / 5

Total

17

/

20

Passed

Description

58%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 concrete, policy-rich description that clearly delineates the skill's advisory contract and single-writer discipline, but it never states when to invoke it (the trigger label and caller contexts live only in the body) and omits the natural "PR" abbreviation. It reads more as an operations contract than a discoverability description.

Suggestions

Add an explicit trigger clause naming the activating label, e.g. "Use when a pull request in microsoft/apm is labelled panel-review and has status/accepted".

Include natural synonyms users would say -- "PR" alongside "pull request", and "code review" / "PR review" alongside "advisory review" -- so the description matches real invocation phrasing.

Trim policy lines ("Do not spawn inactive stubs", "no verdict labels") in favor of one phrase naming when the skill applies, which the description currently lacks.

DimensionReasoningScore

Specificity

The description lists several concrete actions ("run a multi-persona expert advisory review", "selects a surface-gated roster plus a CEO synthesizer", "surfaces findings, prioritizes follow-ups, and renders a ship-recommendation"), but coverage has minor gaps -- the review lenses are never enumerated and part of the budget is spent on policy statements ("Do not spawn inactive stubs") rather than capabilities.

4 / 5

Completeness

The "what" is clear (advisory panel, "ONE recommendation comment, no verdict labels, no merge gating"), but the "when" is only weakly implied by "on a labelled pull request in microsoft/apm" -- there is no "Use when..." clause naming the activating label or the invoking caller. Per the rubric guideline, a missing explicit trigger clause caps completeness at 3.

3 / 5

Trigger Term Quality

Relevant natural terms are present ("pull request", "advisory review", "panel"), but common variations are missing: the abbreviation "PR" never appears, nor do phrases a user would naturally say like "review this PR" or "code review". It sits above anchor 2 (more than generic keywords) and below anchor 4 (keyword coverage is not good, just partial).

3 / 5

Distinctiveness Conflict Risk

The niche is clear (pinned to microsoft/apm, labelled PR, advisory-only contract with "no merge gating"), but the activating label is not named, and closely related sibling skills in the same family (scheduler, merge-worker) fire on the same labelled PRs -- minor overlap risk with closely related skills rather than the minimal risk of anchor 5.

4 / 5

Total

14

/

20

Passed

Validation

75%

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

Validation — 12 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

skill_md_line_count

SKILL.md is long (741 lines); consider splitting into references/ and linking

Warning

frontmatter_unknown_keys

Unknown frontmatter key(s) found; consider removing or moving to metadata

Warning

relative_links

Relative link issues: 10 missing, 10 deeper-than-1-level

Warning

referenced_paths_exist

Referenced path issues: 2 missing

Warning

Total

12

/

16

Passed

Repository

microsoft/apm

Path

packages/autopilot/autopilot-pr-review-worker/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.