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.
A tight, guardrail-rich triage workflow: near-ideal conciseness and validation-checked sequencing for a risky batch-write skill. The main gap is that the GraphQL/project-field read operations — the hardest steps to execute correctly — are described but not given as concrete commands.
Suggestions
Provide the concrete GraphQL query (or gh api command) used to read an issue's native Issue Type field.
Show the command or query for reading Project Priority and Status from Project 199, since 'Read Project Priority and Status from Project 199' currently gives no executable path.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Every line is operational doctrine ("Use labels for routing and action queues", "Do not use another label guide"); there is no concept explanation or padding, and the imperative single-sentence style keeps tokens minimal while preserving clarity. | 5 / 5 |
Actionability | Gives copy-paste-ready gh CLI commands with explicit --json field lists for single and batch modes, plus a concrete write ordering. It falls short of 5 because the two most delicate steps — reading Issue Type "through the GitHub GraphQL API" and Project fields "from Project 199" — provide no command or GraphQL query, and "resolve all live IDs" is asserted rather than shown. | 4 / 5 |
Workflow Clarity | Steps 1-5 are clearly sequenced with explicit validation for a batch write operation: dry-run proposal, explicit apply/skip acceptance, re-read of live state before writes, a stop-and-repropose condition when state differs, and a final report including skipped writes and why. This satisfies the batch-operation validation requirement, so the cap at 3 does not apply. | 5 / 5 |
Progressive Disclosure | The body is a lean overview that delegates canonical policy to four clearly signaled, one-level-deep references in a numbered, purpose-stated list ("The policy package is the source of truth for..."), with nothing bulky inlined. No bundle directories exist inside this skill, and the structure is the ideal overview-plus-references split. | 5 / 5 |
Total | 19 / 20 Passed |