Content
65%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 delivers highly concrete, executable guidance — exact script command, exact Sourcegraph regexes, and precise output formats with a worked URL example — and is appropriately lean. However, two truncated instructions break the workflow mid-sentence, the referenced script is missing from the bundle, the heading hierarchy is inconsistent, and there are no validation checkpoints for batch audits.
Suggestions
Complete the two truncated instructions: the dangling "Follow the" sentence after the Objective and "Do not attempt to search if you do not have the information to..." in section 3 — both currently leave the reader with unfinishable steps.
Bundle the referenced scripts/findDeprecations.mjs in the skill's scripts/ directory (or inline the command's expected behavior), and fix the heading hierarchy so sections 1–4 are all H2.
Add validation/verification steps for comprehensive audits: what to do on zero or ambiguous hits, how to sanity-check query results, and when to re-run or report uncertainty before presenting usage counts as decision input.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes competence — exact queries, a script command, and output formats with no tutorials on concepts Claude already knows. Minor trimming is possible: the objective's stakes sentence ("This information should be as accurate... to inform decisions") is soft padding, and the URL-encoding example is verbose, placing it at the 4 anchor rather than a fully token-optimal 5. | 4 / 5 |
Actionability | Mostly executable guidance: the exact command "yarn node scripts/findDeprecations.mjs --json", copy-paste Sourcegraph regexes with NAME_HERE/PROP_HERE substitution, and a concrete output table spec with a fully worked example URL. It falls short of 5 because two instructions are truncated mid-sentence ("Follow the" and "Do not attempt to search if you do not have the information to"), and the referenced scripts/findDeprecations.mjs is not bundled, leaving minor but real execution gaps. | 4 / 5 |
Workflow Clarity | A numbered sequence is present (scope → deprecation type → Sourcegraph search → structured output) with a useful guard ("Only attempt to search for one deprecation at a time"), but checkpoints are missing or implicit: no verification of search results, no handling of zero/ambiguous hits, and one checkpoint instruction is cut off mid-sentence. A comprehensive audit is also a batch operation, and per the rubric cap, missing validation steps in batch workflows cap this at 3. | 3 / 5 |
Progressive Disclosure | The four sections give the body some structure, but organization is inconsistent (section 2 uses an H3 nested under section 1's H2, and a dangling "Follow the" sentence sits between sections), and the body references paths that are not part of the bundle (scripts/findDeprecations.mjs with no scripts/ directory, plus an external .claude/skills path). References exist but are not clearly signaled against an actual bundle structure, matching the 3 anchor. | 3 / 5 |
Total | 14 / 20 Passed |