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 dense, actionable instruction skill: almost every line is a specific command, flag, limit, or gotcha rather than general knowledge, and install flows include verification steps. The main defects are the vague Hermes/OpenClaw setup pointer and the two referenced files being absent from the provided bundle, which leaves the skill's progressive-disclosure links unresolvable.
Suggestions
Ship TROUBLESHOOTING.md and REFERENCE.md in the skill bundle (e.g., under references/); both are linked from the body but neither file exists in the provided bundle, so the links currently dangle.
For Hermes and OpenClaw, anchor "Follow the matching integration setup instructions" to a concrete path or link so the setup step is executable like the other platforms.
Tighten lines that restate the obvious, e.g. "They send what they record to the local agentmemory server, which stores it on this machine" — the storage behavior is already conveyed by the spool and server notes.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Nearly every line carries plugin-specific facts Claude would not know ("registers 12 hooks", "at most 5 MiB, kept up to 7 days", AGENTMEMORY_* flags, "up to 4,000 characters") with no padding about general concepts. Minor trimming is possible (e.g., "They send what they record to the local agentmemory server, which stores it on this machine"), keeping it at anchor 4 rather than the fully lean anchor 5. | 4 / 5 |
Actionability | Copy-paste-ready commands cover the main platforms ("/plugin marketplace add rohitg00/agentmemory", "devin plugins install rohitg00/agentmemory#plugin", "agentmemory connect pi --with-mcp", "/hooks", "/agentmemory-status", "http://localhost:3113"). Not 5 because the Hermes/OpenClaw branch says only "Follow the matching integration setup instructions" without saying where those instructions live — a concrete gap in executable guidance. | 4 / 5 |
Workflow Clarity | Each install flow is sequenced with explicit verification checkpoints ("Run /hooks inside Devin to confirm the agentmemory hooks are loaded", "verify a real turn creates an observation", "inspect /agentmemory-status") plus an error-recovery path ("If observations are missing, confirm the plugin is enabled and the server is running"). Anchor 5 is not reached because the fix-retry loop is delegated to a TROUBLESHOOTING.md that is not in the bundle and the Hermes/OpenClaw sequence is underspecified. | 4 / 5 |
Progressive Disclosure | The pattern is right: a concise overview with clearly signaled one-level references ("See [TROUBLESHOOTING.md](TROUBLESHOOTING.md)", "The exact registered hook events live in REFERENCE.md, generated from plugin/hooks/hooks.json") and a "See also" section for sibling skills. Scored against the actual bundle, however, neither referenced file exists (no references/, scripts/, or assets/ directories and no TROUBLESHOOTING.md or REFERENCE.md alongside SKILL.md), so the best links dangle — a minor organization gap per anchor 4 rather than a well-resolved anchor 5. | 4 / 5 |
Total | 16 / 20 Passed |