Content
85%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.
An exceptionally actionable and well-sequenced operations contract with excellent progressive disclosure — every command is executable, verification is built into every step, and detail is properly split across an intact reference bundle. The main drag is conciseness: the checked_gh wrapper mechanics and an incident-attribution aside add substantial length that a competent reader doesn't need inline.
Suggestions
Move the checked_gh.py internal behavior details (environment scrubbing, host-operand conflict rules, rejected command classes) into the script's docstring or a short reference file, keeping only the invocation syntax and the actor-check guarantee in SKILL.md.
Trim or relocate the incident-attribution paragraph about the 'ChatGPT execution account' — it addresses a different tool's incident forensics and adds context cost without supporting this skill's operating contract.
Consolidate the repeated identity/recheck rules in step 2 ('Switching account, credential, host or interface invalidates earlier identity evidence. Recheck before the next write; do not reuse a prior session's pass') with the earlier bind-identity instructions to remove redundancy.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body avoids explaining basic concepts and is dense with operational rules, but sections could be tightened: step 2 devotes ~40 lines to the checked_gh wrapper's internal behavior ("removes inherited GH_REPO and alternate token variables from the child, and rejects conflicting host operands") that belongs in the script's own documentation, and the incident-attribution paragraph ("the ChatGPT execution account") is off-topic for this skill's core contract. This lands at 'mostly efficient but some unnecessary explanation or could be tightened' rather than the minor-trim level of 4. | 3 / 5 |
Actionability | Guidance is fully executable: concrete commands with exact flags and --json field lists ("gh repo view HOST/OWNER/REPO --json nameWithOwner,visibility,isPrivate,viewerPermission,url"), copy-paste-ready checked_gh.py invocations with real argument shapes, a fill-in delta-preview template, and a ready-to-run read-only quick reference. Specific evidence requirements per mutation type (e.g. "Secret/variable change | Metadata and consumer behavior, never secret value disclosure") cover the common cases. | 5 / 5 |
Workflow Clarity | The universal operating contract is an explicitly sequenced 8-step process (classify → bind identity → preview delta → choose interface → mutate once → independent readback → report outcome → scoped authentication) with strong validation checkpoints and feedback loops: "After a timeout or 5xx, read back first to determine whether the first request landed", "If readback differs, report failed/no-op or partially applied… and keep recovery available", and bounded polling with a pending state. This matches the anchor for clear sequence with explicit validation and error-recovery loops, including for the destructive/batch operations this skill covers. | 5 / 5 |
Progressive Disclosure | SKILL.md is a well-structured overview with a task→reference routing table pointing to 11 one-level-deep reference files (all of which exist in references/), a bundled script clearly linked, and an anchor link (references/pr_operations.md#merging-prs) that resolves to a real '### Merging PRs' section. Operation-specific detail is appropriately deferred while the cross-cutting contract stays inline; navigation by task is easy. | 5 / 5 |
Total | 18 / 20 Passed |