Content
92%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is an exemplary procedural skill: lean, fully executable, with a clearly sequenced workflow and robust validation/feedback loops (scrub-retry, yes-gate, re-review of changed drafts). Its one real defect is bundle integrity — three of the six referenced paths (kinds/, issue.md, comment.md) do not exist in the shipped bundle, leaving the detection and drafting steps without their promised inputs.
Suggestions
Ship the missing `kinds/` files referenced by the Detect section ("Read every file in `${CLAUDE_SKILL_DIR}/kinds/`"), or inline the kind definitions if the directory is not part of the bundle.
Add the missing `issue.md` and `comment.md` templates referenced by the Draft section, or replace the references with the layout content inline.
Consider referencing `scripts/scrub.test.sh` (present but unreferenced) from the body so the bundle's test coverage is discoverable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~63-line body is pure procedure with zero padding and no explanations of concepts Claude already knows — every line instructs ("Run `${CLAUDE_SKILL_DIR}/scripts/read.sh <id>`... It fails → show its error and stop"). Matches 'lean and efficient; assumes Claude's competence; every token earns its place'; not 4 because the over-explanation-that-could-be-trimmed in that anchor is entirely absent. | 5 / 5 |
Actionability | Fully executable, copy-paste-ready commands throughout: `read.sh <id> <turn>`, `past.sh`, `gh issue view <N> -R saadshahd/moo.md --comments`, `mktemp`, `scrub.sh <id> <draft>`, and the complete issue pipeline `tail -n +2 <draft> | gh issue create -R saadshahd/moo.md --label harvest --title "$(sed -n '1s/^Title: //p' <draft>)" --body-file -`, plus a concrete five-row decision table. Matches 'fully executable; copy-paste ready'; not 4 because there are no gaps in the commands — every placeholder (<id>, <turn>, <draft>, <N>) is defined. | 5 / 5 |
Workflow Clarity | A clear nine-phase sequence (Read → Detect → Name the opportunity → Value → Match past issues → Draft → Scrub → Show → File on yes) with explicit validation and feedback loops: "It fails → show its error and stop", "Exit 1 → replace each term it lists with a placeholder and run it again", "Never show a draft that has not passed", the reporter's yes-gate, and "A draft changed after the yes goes back through Scrub and Show". The outward-facing filing operations carry full validation, so the missing-validation cap does not apply; matches the explicit-validation-with-feedback-loops anchor. | 5 / 5 |
Progressive Disclosure | Section structure is clean and references are one level deep, but scored against the actual bundle: of the six referenced paths, `scripts/read.sh`, `scripts/past.sh` and `scripts/scrub.sh` exist while `${CLAUDE_SKILL_DIR}/kinds/`, `${CLAUDE_SKILL_DIR}/issue.md` and `${CLAUDE_SKILL_DIR}/comment.md` are absent from the bundle. The Detect step ("Read every file in `${CLAUDE_SKILL_DIR}/kinds/`") and the Draft templates depend entirely on files that are not shipped, so navigation fails exactly where detail is deferred. Missing half the referenced paths is beyond anchor 4's 'minor organization gaps'; anchor 3 fits the impact. | 3 / 5 |
Total | 18 / 20 Passed |