Content
78%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.
A lean, highly operational instruction body with an exact, machine-readable output contract, weakened by a dangling bundle reference: the entire 'What to look for' section delegates to lens.md, which is not present, so the skill's core review criteria never load. Token efficiency and in-body structure are excellent; the bundle just fails to deliver the referenced detail file.
Suggestions
Ship lens.md in the bundle (e.g. as references/lens.md) and update the pointer to 'Read references/lens.md in this skill's directory and apply it' — the body's only reference currently dangles, so the review criteria are unavailable.
Add a one-line fallback for the missing-lens case, such as 'If lens.md is absent, check data loss, concurrency and retry handling, error paths, and regressions in existing behavior', so the skill remains executable standalone.
Add an explicit pre-report checkpoint to the workflow, e.g. 'Before writing the report, confirm each finding names a concrete input, state, or timing that reaches it', to close the validation gap.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Every section carries operational instruction ('Read the change with `git -C <checkout> diff <base> <head>`', 'For a PR, use the merge base of the PR head and its base branch, never the current tip of the base branch', 'The last line must be `Rating: <value>` with nothing after it, so a script can read it'). No explanations of concepts Claude already knows and no padding; not the 4 anchor because there is nothing left to trim. | 5 / 5 |
Actionability | Concrete commands ('git -C <checkout> diff <base> <head>' with the 'Add `-- <path>` to read one file' variant) and a copy-paste-ready output template with severity levels, file:line locations, and a machine-readable rating line. Not 5 because the core guidance 'Read `lens.md` in this skill's directory and apply it' points to a file absent from the bundle, leaving the actual review criteria unspecified; not 3 because everything written in the body is executable and concrete, not vague or pseudocode. | 4 / 5 |
Workflow Clarity | Clear sequence: gather inputs with an explicit fallback ('If the caller leaves one out, work it out and say what you used'), read the diff with a specific command, apply the lens, then emit the exact report shape ending in a parseable rating line. Not 5 because the apply-lens step dead-ends at a missing file and there is no explicit self-check checkpoint before writing the report; not 3 because the sequence and output contract are fully defined. The skill is read-only, so the destructive/batch validation cap does not apply. | 4 / 5 |
Progressive Disclosure | The body is well sectioned (Inputs, What to look for, Rules, Output format) and its single reference is clearly signaled and one level deep ('Read `lens.md` in this skill's directory and apply it'), but the bundle contains no lens.md — the referenced content is neither inline nor present, so the disclosure structure is broken in practice. Not 4 because a dangling reference to the skill's core content is more than a minor organization gap; not 2 because the body itself is well organized and nothing is buried or inlined that belongs elsewhere. | 3 / 5 |
Total | 16 / 20 Passed |