Content
81%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 highly actionable with copy-paste-ready commands, explicit defaults, output formats, and concrete failure recovery, and the single-action workflow is unambiguous. The main weakness is redundancy — repeated dedup and anti-put guidance, duplicated trigger lists, and inline version-specific details that could be tightened.
Suggestions
State the 24h content-hash dedup once (Contract or Defaults, not both) and merge the intro's "don't reach for gbrain put or commit-then-sync" into the Anti-Patterns section to remove the repetition.
Drop the "When to invoke" section or reduce it to a pointer, since it restates the frontmatter triggers verbatim.
Move version-sensitive specifics ("v0.38 write-through plumbing") out of the main flow or into a changelog/deprecated note so the guidance ages better.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient, but the 24h dedup sentence appears verbatim in both Contract and Defaults, "don't reach for gbrain put" is stated in both the intro and Anti-Patterns, "When to invoke" duplicates the frontmatter triggers, and version-sensitive details ("v0.38 write-through plumbing", date-stamped receipt example) are not isolated — more than minor trimming is needed. | 3 / 5 |
Actionability | Commands are copy-paste ready and cover every input mode (inline, --file, --stdin) and flag (--slug, --type, --source, --quiet, --json), the exact receipt output format is shown, and failure handling gives per-exit-code recovery actions (poll gbrain write-request, never re-capture). | 5 / 5 |
Workflow Clarity | This is a simple single-action skill with an unambiguous invocation (run gbrain capture with the content), and the "When it fails" section supplies explicit error-recovery guidance; it is not a destructive or batch operation, so no validation cap applies. | 5 / 5 |
Progressive Disclosure | Sections are well-organized with clearly signaled one-level references to sibling skills and an operator protocol, but the referenced paths (../conventions/agent-operator-protocol.md, skills/*/SKILL.md) do not exist in this bundle, and at ~100 lines the body sits above the under-50-line simple-skill exception. | 4 / 5 |
Total | 17 / 20 Passed |