Conciseness

The ~40-line body is lean: no concept explanations Claude already knows, and every sentence adds non-obvious information ("This runs hybrid BM25 plus vector plus graph-expanded search", "REST handlers whitelist fields and drop unknown ones"). The tool-family lists are dense indexes, not padded prose — every token earns its place.

Actionability

The Quick start gives concrete tool calls with parameter names ("memory_save with content, concepts, files"; "memory_smart_search with query and limit"), and the Workflow picks specific tools per task — but exact/required parameters are deferred to REFERENCE.md, which is not present in the bundle, so guidance is not fully executable standalone. That is a minor-to-real gap placing it below the fully-executable 5 anchor and above the incomplete 3 anchor.

Workflow Clarity

The Workflow section is a clear 3-step sequence (pick the narrowest tool → look up exact parameters in REFERENCE.md → pass only documented fields) with a pre-flight check in step 2 and a failure-mode warning (unknown fields are dropped). No explicit validation or feedback loop exists, and while the skill indexes destructive tools (memory_governance_delete, memory_heal), its own workflow is selection/lookup rather than executing them — so the destructive-cap doesn't strictly apply, but the missing error-recovery keeps it below 5.

Progressive Disclosure

Structure is good — well-organized sections (Quick start, Tool families, Workflow, See also, Reference) with the bulky tool table correctly split out to a one-level-deep, clearly signaled REFERENCE.md. However, scored against the actual bundle: no references/, scripts/, or assets/ directories exist and REFERENCE.md is absent, so the skill's central deferred resource ("Full tool table with parameters... lives in REFERENCE.md") is unreachable. A dead primary reference is more than the 'minor organization gaps' of the 4 anchor, though the in-body structure is far better than the 2 anchor's inlined bulk.

Specificity

"Map of every agentmemory MCP tool, what each does, and its parameters" names the domain and several concrete deliverables (tool map, per-tool descriptions, parameter reference), with only minor gaps (no mention of the tool families or count). It lists what the artifact contains rather than several distinct action verbs, so it sits between the 3 and 5 anchors — several specific items but not comprehensive action coverage.

Completeness

The 'what' is explicit ("Map of every agentmemory MCP tool, what each does, and its parameters") and the 'when' is explicit with concrete trigger clauses ("Use when choosing which memory tool to call, when a tool name or argument is unclear, or when answering what agentmemory can do via MCP"), matching the top anchor's pattern of a clear what plus multiple concrete when-triggers.

Trigger Term Quality

Natural trigger phrases like "choosing which memory tool to call", "when a tool name or argument is unclear", and "what agentmemory can do via MCP" match how a user would phrase the need. Coverage is good but slightly narrow — common variants like specific tool names (e.g. memory_save, memory_recall) or generic phrasings like "memory search"/"remember" are absent — so it misses the comprehensive-synonyms bar of the 5 anchor without being merely generic.

Distinctiveness Conflict Risk

The niche is tightly scoped to agentmemory MCP tools with trigger clauses keyed to tool/argument ambiguity, so it is unlikely to fire for unrelated memory or PDF/document skills. It carves a clear boundary distinct from sibling agentmemory skills (REST API, config) rather than overlapping them.