Content
71%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 a lean, well-sectioned overview with genuinely concrete tool-call guidance, sensible destructive-operation warnings, and host-specific integration detail Claude would not know. Its one real defect is that the load-bearing reference to REFERENCE.md — where exact parameters are supposed to live — points at a file absent from the bundle, which both weakens progressive disclosure and leaves parameter lookups unsatisfiable.
Suggestions
Ship REFERENCE.md in the bundle (e.g., references/REFERENCE.md) or inline the required-parameter table for the core tools, since Workflow step 2 and the Reference section both depend on a file that does not exist.
Tighten the intro's host-integration sentences (Hermes/OpenClaw/Pi/OMP) into a short table or move them to the reference file so the overview stays lean.
Add a one-line validation habit after tool calls beyond memory_recall (e.g., checking the response for dropped/unknown fields), turning the existing budget-clipping recovery step into a general checkpoint.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and operational — tool names with parameters ("`memory_save` with `content` ... `concepts` ... `files`"), no padding, and no explanation of concepts Claude already knows. The intro paragraph packs host-specific integration detail (Hermes, OpenClaw, Pi, OMP prefixes and `--with-mcp`) that could be tightened or split out, keeping it at anchor 4 rather than the fully lean anchor 5. | 4 / 5 |
Actionability | The Quick start gives concrete, executable calls with real parameter names and values ("`memory_smart_search` with `query`, `limit`, and `targetLayer: \"memory\"`"), and Workflow step 4 gives specific handling for `matched_count: 0` vs `excluded_by_budget > 0` with the `expandIds` follow-up. It falls short of anchor 5 because exact required-parameter lookup for every tool is deferred to REFERENCE.md rather than shown, leaving minor gaps. | 4 / 5 |
Workflow Clarity | The four-step Workflow (pick narrowest tool → check parameters → pass only documented fields → interpret budget results) is clearly sequenced, and the destructive-tools section includes an explicit checkpoint ("Deletion cannot be undone; confirm the IDs with the user first") plus a recovery path for clipped results. It lacks the full validate-fix-retry loops and checklists of anchor 5, so anchor 4 fits. | 4 / 5 |
Progressive Disclosure | Sections are well-organized and the design intent is right — a lean overview deferring the "Full tool table with parameters" to REFERENCE.md one level deep — but REFERENCE.md does not exist in the bundle (no references/ directory and no REFERENCE.md anywhere), so the skill's primary navigation target is a dangling reference. Structure is good yet the broken reference undermines navigation more than the 'minor organization gaps' of anchor 4. | 3 / 5 |
Total | 15 / 20 Passed |