Conciseness

The body is efficient and assumes competence — it never explains MCP, worldbuilding, or novel-writing basics — and is dense with domain-specific guidance. Minor instances could be trimmed: `author.record_decision`/`author.defeat` policy and retcon guidance each appear in both "Authoring loop" and "Core tools", and "Contract 2.0.0 is the served contract" pins a version number inline, which the guidelines penalize outside a compatibility section. Not anchor 5 because not every token earns its place; not anchor 3 because there is no unnecessary concept explanation.

Actionability

Provides mostly executable guidance: concrete tool names, the registration command `deloreis mcp --project <directory>`, citation JSON like `{"kind":"cited","decisions":["d-0001"],"state":"ratified"}`, and specific parameters (`expected_digest`, `layer: "certain"`). Minor gaps remain — exact tool schemas are deferred to `tools/list` and the complete first walkthrough to references/first-session.md — so it falls short of anchor 5's fully copy-paste-ready coverage of common cases.

Workflow Clarity

The five-step "Authoring loop" is clearly sequenced (open → typed operations → decisions → inspect → materialize/discard) with explicit validation checkpoints: "Check the returned findings even if the call itself succeeded", "`author.materialize`... It refuses errors", digest-freshness handling, and "Unknown citations are errors". Feedback loops for error recovery are present (repair mechanical errors, `author.discard`, escalate conflicts to the user's rulings), matching anchor 5 rather than anchor 4's "minor validation gaps".

Progressive Disclosure

Good structure with a real, one-level-deep reference (references/first-session.md) that is well-signaled via "[First session](references/first-session.md)" and an anchor link that resolves, and installation detail is appropriately deferred to it. Minor organization gaps keep it below anchor 5: the 29-tool catalog is inlined in the body rather than referenced out, and some guidance (retcon/defeat, record_decision policy) is duplicated across "Authoring loop" and "Core tools" instead of consolidated.

Specificity

Names several concrete actions — "Record and query persistent lore through local MCP tools", "checks on generated additions to a fictional world", "maintaining lore" — with a named mechanism, but coverage of what recording and checking involve has minor gaps. It exceeds anchor 3 (only 1-2 concrete actions) yet is not comprehensive enough for anchor 5.

Completeness

Explicitly answers both what ("Record and query persistent lore through local MCP tools", "Keep fiction and worldbuilding consistent") and when ("Use when maintaining lore for AI-assisted novels, series bibles, campaigns, competing story ideas, retcons, or checks on generated additions") with concrete trigger phrases. This clearly matches the anchor-5 example structure rather than anchor 4, where the 'when' would be less specific.

Trigger Term Quality

Includes natural terms users would say — "AI-assisted novels, series bibles, campaigns, competing story ideas, retcons" — plus explicit name variants "De Lore Is, DLI, or deloreis". A few common synonyms are missing (e.g., "canon", "continuity", "story bible", or the ".dli" file extension), keeping it below anchor 5's comprehensive coverage.

Distinctiveness Conflict Risk

Occupies a clear niche (fiction/worldbuilding lore consistency) with distinct triggers such as "retcons", "series bibles", and "competing story ideas", plus explicit project-name triggers ("De Lore Is, DLI, or deloreis"). Conflict risk with other skills is minimal; anchor 4's "minor overlap risk" would require broader, less distinctive triggers.