Content
88%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 an exemplary single-command skill reference: complete command surface, concrete defaults, an exact receipt format, error-code remediation with recovery loops, and clear negative guidance (anti-patterns, when NOT to use) that routes bulk and specialized ingestion to the right tools. Only minor duplication between Contract/Defaults and intro/Anti-Patterns keeps it from full marks on token efficiency.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and actionable with no concept explanations, but has minor repetition that could be trimmed: the positioning of `capture` over `gbrain put` appears in both the intro ("Don't reach for `gbrain put`") and Anti-Patterns, and the daemon's 24h content-hash dedup is stated identically in both Contract ("The daemon's 24h content-hash dedup catches re-captures") and Defaults. Fits anchor 4 ("Efficient; minor instances of over-explanation that could be trimmed") rather than anchor 5's every-token-earns-its-place. | 4 / 5 |
Actionability | Fully executable guidance throughout: copy-paste-ready commands covering all input modes and flags (inline, --file, --stdin, --slug, --type, --quiet, --json), concrete defaults (slug format, `note` type, 80-char title cap, frontmatter stamps), an exact example output receipt, and specific exit-code remediation ("poll `gbrain write-request <request_id>`; never capture it a second time"). Matches anchor 5's copy-paste-ready coverage of common cases. | 5 / 5 |
Workflow Clarity | This is a simple, single-purpose skill whose one action (run `gbrain capture`) is unambiguous, satisfying the simple-skill exception, and it goes further with an explicit error-recovery loop: exit 10 → tell the user it's saved and poll rather than re-capture, refusal → report what was refused without sneaking past the scanner, MCP scope errors → name the source needing a grant. It is not a destructive or batch operation (bulk looping is explicitly an anti-pattern), so the validation cap does not apply. | 5 / 5 |
Progressive Disclosure | Sections are well-organized (Contract, When to invoke, How to use, Defaults, Output Format, When it fails, Anti-Patterns, When NOT to use) and the cross-references (../conventions/agent-operator-protocol.md and sibling skill SKILL.md paths) are clearly signaled one level deep. It is not anchor 5: the skill is ~100 lines all inline with no content split, and the referenced paths point outside the skill directory (no bundle files exist here to verify), so structure falls to "good structure; most content appropriately placed" rather than a clearly split, easily navigable overview. | 4 / 5 |
Total | 18 / 20 Passed |