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.
A well-engineered operational skill: dense, imperative, with concrete commands, explicit user-consent checkpoints, verification loops, and a clean split of platform detail into five real reference files. The main gaps are minor — a version-specific config note in the main flow and one step that defers its command to an external skill's reference.
Suggestions
Move the "Since v0.4.11" .memsearch.toml note into a short 'Version notes / old patterns' section (or drop the version number and just state the current rule) so time-sensitive detail doesn't sit in the main flow.
Inline the exact transcript-drill command (or add a references file in this bundle containing it) instead of pointing to the memory-recall skill's platform reference, so flow C is executable without leaving the skill.
Trim the explanatory sentences in section C (e.g. the comparison with the background pass at the end) to tighten the body further.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and directive — numbered stages, terse intent routing, no explanation of concepts Claude already knows; every section carries operational weight. Not a 5: the time-sensitive "Since v0.4.11, project-local .memsearch.toml accepts only allowlisted local indexing keys" note sits in the main flow rather than a versioned/deprecated section, and a few sentences (e.g. the closing rationale in C) could be trimmed. | 4 / 5 |
Actionability | Mostly copy-paste ready: `printf ... | memsearch skills add --name ... --body-file -`, `memsearch skills status/list/install`, exact `config get/set` keys, and a complete `git -C .memsearch/skill-candidates log --oneline -5 2>/dev/null || true`. Not a 5: the transcript-drill step in C defers the actual command to an external skill's reference ("see the memory-recall skill's platform reference for the exact command"), so that step is not executable from this skill alone. | 4 / 5 |
Workflow Clarity | Stages 0→1→2 are explicit, intent routing maps requests to flows, and each flow has explicit validation checkpoints and error-recovery loops: verify uncertain details before drafting ("re-check it against the source... or keep that step general"), hash-based `skills status` to detect evolved candidates, skim-and-re-check before install, confirm install destination, and a fallback in C ("Only if that command fails (unknown format) fall back to reading the raw file directly"). The install operation is gated on explicit user confirmation, so the destructive/batch cap does not apply. | 5 / 5 |
Progressive Disclosure | Clear overview with well-signaled, one-level-deep references: all five platform files (`references/claude-code.md`, `codex.md`, `openclaw.md`, `opencode.md`, `dsh.md`) exist in the bundle and are each named with their platform in the routing list; platform-specific config prefixes and install paths are correctly split out of the main body. No nested references, and the body stays at overview altitude. | 5 / 5 |
Total | 18 / 20 Passed |