Conciseness

The body is lean and assumes Claude's competence: no explanations of what MCP or REST are, no filler, every line carries operational information ('Windows: use WSL2. Native Windows runs the server but `connect` is not supported there.'). This matches the anchor-5 'every token earns its place' example; there is nothing to trim without losing information.

Actionability

Concrete, copy-paste-ready command ('agentmemory connect claude-code # or cursor, codex, gemini-cli, ...'), a concrete reload example ('/mcp' in Claude Code), and a numeric failure signature ('Only 7 tools means the MCP shim could not reach a server'). It stops short of 5 because the verification step ('show the full tool set with a server running') doesn't specify how to check the server is running, and the adapter-name table the workflow depends on is delegated to REFERENCE.md.

Workflow Clarity

The workflow has a clear detect -> connect -> verify sequence with an explicit validation checkpoint (step 3) and an error-recovery pointer for a named failure mode. It sits at 4 rather than 5 because the validate -> fix -> retry loop is only pointed at ('see ../_shared/TROUBLESHOOTING.md') rather than spelled out, and the check in the verification step is partially implicit; it is above 3 because validation is explicit and tied to a concrete failure signature.

Progressive Disclosure

The body itself is well-sectioned and its references are clearly signaled and one level deep ('The full adapter list ... lives in REFERENCE.md'). However, the bundle contains no references/ directory or REFERENCE.md, and the referenced ../_shared/TROUBLESHOOTING.md is also absent, so the primary references the workflow depends on are dangling relative to the actual bundle structure. Not 4 because a navigation path the reader cannot complete is more than a minor organization gap; not 2 because the in-body structure is good and references are prominent rather than buried.

Specificity

The description names the domain ('wires into host coding agents via the connect command') and lists several concrete actions/scenarios ('installing agentmemory into a specific agent', 'when asked which agents are supported', 'when a connect adapter writes the wrong config path'). It is not a 5 because the 'what' is a single mechanism (wiring via connect) and coverage of what the wiring actually does (merge server config, preserve existing servers) is left implicit.

Completeness

It explicitly answers both questions: the 'what' ('How agentmemory wires into host coding agents via the connect command') and an explicit 'Use when...' clause with three concrete trigger situations. This matches the anchor-5 example structure of clear what plus concrete trigger phrases; it is not below 4 because the 'when' is fully explicit rather than merely present.

Trigger Term Quality

Phrases like 'installing agentmemory into a specific agent', 'which agents are supported', and 'connect adapter writes the wrong config path' are natural user utterances for this niche. A few natural variations are missing (e.g., 'set up', 'configure agentmemory for claude code/cursor'), so it falls just short of the comprehensive-synonym anchor at 5.

Distinctiveness Conflict Risk

The trigger set is tightly scoped to connect/adapter installation issues, clearly distinguishing it from sibling skills (mcp-tools, rest-api, hooks) and from generic agent-config skills. Minimal conflict risk with 'use when' conditions naming distinct failure modes.