Content
78%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 efficient, highly actionable connector guide with concrete commands, host-specific flags, and a workflow that includes authorization and validation checkpoints. Its main weaknesses are the dangling references to REFERENCE.md and TROUBLESHOOTING.md, which are absent from the bundle and break the progressive-disclosure structure the workflow depends on, plus minor time-sensitive detail (a pinned npx version) in the Notes.
Suggestions
Ship REFERENCE.md (with the adapter table) and TROUBLESHOOTING.md in the skill bundle — Workflow step 2 directs the reader to "the table in REFERENCE.md", which is currently missing and blocks execution for unlisted hosts.
Move the pinned "npx skills@1.7.1" version and the provenance note "generated from src/cli/connect/" out of the body (e.g., into REFERENCE.md or a maintenance section) to keep the main file free of time-sensitive tokens.
Add an explicit recovery loop to Workflow step 3 (what to do when the save/recall check fails or the host still doesn't list agentmemory's tools) instead of relying solely on the TROUBLESHOOTING.md link.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and operational with no concept explanations Claude already knows, but the pinned "npx skills@1.7.1" is time-sensitive version information outside any deprecated/old-patterns section, and "generated from `src/cli/connect/`" is repo-internal detail that costs tokens without aiding execution. Not 5 because of these trimmable tokens; not 3 because padding is minimal and isolated. | 4 / 5 |
Actionability | Fully executable, copy-paste-ready commands cover the common cases: "agentmemory connect claude-code", "agentmemory connect pi --with-mcp", "agentmemory connect dsh --with-hooks", "devin plugins install rohitg00/agentmemory#plugin", "npx skills@1.7.1 add rohitg00/agentmemory --agent hermes-agent --skill '*'", plus concrete reload/verification commands (/reload, /agentmemory-status, /mcp). | 5 / 5 |
Workflow Clarity | The 3-step Workflow has clear sequencing with an authorization guard ("If unknown, ask which host to configure before changing files"), an explicit failure signal ("an adapter that prints manual setup instructions has not completed installation"), and validation ("inspect its actual tool inventory, and perform an authorized save/recall check"). Not 5 because there is no stated recovery path if the save/recall check fails — recovery is delegated to a TROUBLESHOOTING.md link rather than a feedback loop. | 4 / 5 |
Progressive Disclosure | Sections are well organized and the references are clearly signaled and one level deep ("[TROUBLESHOOTING.md](TROUBLESHOOTING.md)", "lives in REFERENCE.md"), but the bundle contains no files at all — neither REFERENCE.md nor TROUBLESHOOTING.md exists — so Workflow step 2's required adapter-name table is unreachable. Not 4 because a missing critical reference is more than a minor organization gap; not 2 because the inline content is appropriately placed and well structured for the main level. | 3 / 5 |
Total | 16 / 20 Passed |