Content
75%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 highly actionable command reference with real, copy-paste-ready invocations, sensible validation for the destructive operation, and clean sectioning. Its weaknesses are repetition of the evidence-vs-hypothesis framing, some decorative prose, and no use of progressive disclosure to move backend internals and full response schemas out of the main file.
Suggestions
State the consult-evidence / surface-hypothesis distinction once (in 'What It Is' or under the commands) and cut the repetitions in the consult section, 'Interpreting Associations', and 'When to Use'.
Move the full JSON response examples, storage architecture diagram, and backend internals into a references/ file (e.g. BACKENDS.md) and keep SKILL.md as a lean overview with one-level-deep links.
For the destructive `forget` command, make the checkpoint explicit: recommend running with --dry-run first and only executing the delete after reviewing the output.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body teaches no Claude-known concepts and most content is skill-specific, but it restates the consult-as-evidence / surface-as-hypothesis distinction three times (in 'What It Is', in the consult section, and again under 'Interpreting Associations'), the ASCII backend diagram duplicates the backend-detection bullets below it, and the epigraph/'filing cabinets' prose is padding. Anchor 3 (mostly efficient but could be tightened) fits: more than minor trimming needed for a 4, but far from the padded anchor-2 pattern. | 3 / 5 |
Actionability | Every command has copy-paste-ready invocations ("python3 scripts/inland_empire.py remember ..." — path verified to exist), with flags (--depth, --type, --before, --dry-run) and example JSON responses covering the common cases. Fully executable with no gaps — anchor 5. | 5 / 5 |
Workflow Clarity | Each command's usage is unambiguous and the destructive `forget` operation does document a validation mechanism (--dry-run), plus error/partial-flag semantics for recovery. It falls short of anchor 5 because the doc never explicitly recommends dry-run-before-delete as a step for destructive or batch forgets — a minor checkpoint gap characteristic of anchor 4. | 4 / 5 |
Progressive Disclosure | Well-organized sections with a single referenced path (scripts/inland_empire.py) that is a real bundle file, and no nested or buried references. But it is a ~263-line monolith: response schemas, storage architecture, and backend internals are inlined rather than split into reference files (the bundle even ships mem0/memory_graph/memory_libsql docs that are never surfaced). Anchor 4 (good structure, minor organization gaps) rather than 5. | 4 / 5 |
Total | 16 / 20 Passed |