evaluate-pr

Evaluate a PR the harness produced — walk the change, run the system together, and build a firm understanding before you merge it. Use after the harness hands a converged PR to you for review (the "Ready for your review" comment), or any time you want to deeply review an agent-authored PR. The human-attentive skill at the back of the chain; the mirror of /intent. Outcomes — merge, close, or fix-it-yourself-and-push - no handing work back to the loop.

Quality

82%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Quality

Content

77%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 content is a well-structured, actionable guide with a clear sequenced workflow, validation gates for destructive operations, and clean progressive disclosure into two reference files. Its main weakness is verbosity from restating the same E-numbered invariants across multiple sections.

Suggestions

Reduce repetition: the E1–E9 philosophy points are restated in the steps and again in 'Hard nevers' — consolidate to one canonical location and reference it.

Trim the philosophy section to the invariants that change behavior; several points restate the same 'you decide, you act / don't hand back to the loop' idea in different words.

Make the bootstrap step slightly more concrete or defer fully to verdict.md rather than leaving 'e.g. ./scripts/bootstrap-worktree.sh . if present, else the project's install' inline.

Dimension	Reasoning	Score
Conciseness	The body is mostly efficient and assumes Claude's competence (no basic-concept padding), but the philosophy section (E1–E9) and repeated restatements of the same invariants across 'How to run', steps, and 'Hard nevers' add noticeable length that could be tightened.	3 / 5
Actionability	Provides concrete executable commands (git fetch, checkout --detach, gh pr diff, git push origin HEAD:feature/<f>) and specific file-reading order, with only minor gaps where guidance stays abstract (e.g. 'bootstrap so the app runs').	4 / 5
Workflow Clarity	An explicit Step 0–8 sequence with a soft understanding-gate checkpoint (Step 6), explicit go-ahead gates before destructive merge/close actions, and clear feedback loops (fix → push → reviewer re-runs → merge), satisfying the validation requirement for destructive operations.	5 / 5
Progressive Disclosure	Clear overview structure with well-signaled one-level-deep references to references/walkthrough.md and references/verdict.md (both verified to exist), each labeled with its purpose and a 'hackable seam' note; content is appropriately split rather than inlined.	5 / 5
	Total	17 / 20 Passed

Description

87%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.

The description is strong: it clearly states what the skill does, when to use it, and its distinct place in the skill chain. Its only weakness is mild verbosity and a couple of missing natural synonyms in the trigger terms.

Suggestions

Tighten the description — phrases like 'The human-attentive skill at the back of the chain; the mirror of /intent' add length without improving trigger discovery.

Add a few common trigger synonyms (e.g. 'review a PR', 'merge an agent PR') users might naturally say.

Ensure third-person voice is consistent throughout; the current phrasing is largely imperative/gerund which reads well, but trim any second-person framing.

Dimension	Reasoning	Score
Specificity	Names the domain (PR review) and several concrete actions ('walk the change', 'run the system together', 'merge', 'close', 'fix-it-yourself-and-push'), with minor gaps in coverage rather than a full action enumeration.	4 / 5
Completeness	Explicitly answers both 'what' (walk/run/build understanding → merge, close, or fix-and-push) and 'when' ('Use after the harness hands a converged PR... or any time you want to deeply review an agent-authored PR') with concrete trigger phrases.	5 / 5
Trigger Term Quality	Includes natural trigger phrases a user would say ('review', 'Ready for your review', 'deeply review', 'agent-authored PR') with good keyword coverage, though it lacks common synonyms like 'merge a PR'.	4 / 5
Distinctiveness Conflict Risk	Occupies a clear niche (agent-authored PR evaluation in a human-in-the-loop chain, the mirror of /intent) with distinct triggers, giving minimal conflict risk with other skills.	5 / 5
	Total	18 / 20 Passed

Validation

100%

Warnings & errors only

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: tdg-ninja/context-specs-factory-ai
Path: skills/human-loop/evaluate-pr/SKILL.md
Commit: 3a7a725

Reviewed: about 19 hours ago

Table of Contents

Discovery Implementation Validation

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.