Conciseness

The body is all procedure with no concept-explanation padding, but key rules are stated repeatedly: the crash-safe persist rule appears in near-identical prose in "Persist as you go", Step 4, and Step 5; supervised/headless guidance repeats beyond the mode table in Steps 1, 2, and 8; and the Non-goals section restates six rules already enforced inline ("An article without preserved sources is opinion" appears twice). Not a 2: the bulk is lean, actionable instruction rather than padded explanation; not a 4: several sections are demonstrably trimmable to a single statement of each rule.

Actionability

Fully executable guidance throughout: exact commands (`exec("grep -rln <topic-keyword> <content-dir>")`, `audit({ path: "<doc-path>.md" })`), a copy-paste TaskCreate list, a complete scoping-rubric template, exact frontmatter YAML, a full article skeleton, and a concrete recap template. Not a 4: the templates and commands cover the common cases end-to-end with no missing key details.

Workflow Clarity

A 9-step mandatory execution order with hard gates, task checkpoints (Step 0), a routing decision table (Step 1), a validation checklist (Step 7) with an audit-then-fix-every-finding feedback loop, crash-resume recovery in Step 5, and explicit Path B/C and headless/supervised variants. Not a 4: validation steps, error-recovery loops, and checklists are all explicitly present, matching the top anchor.

Progressive Disclosure

The body is well-sectioned (one section per step, clear headers), and the ingest procedure is properly delegated to the platform skill — "it lives in the platform `/open-knowledge` skill at `references/ingest-and-sources.md`, not in this pack" — a clearly signaled, one-level-deep reference; no bundle files of its own exist to verify. Not a 5: the ~370-line monolith inlines content (templates, routing scripts, mode table) that could be split into reference files for a lighter overview; not a 3: the single reference is clearly signaled and the structure is consistently well-organized.

Specificity

Lists multiple specific concrete actions covering the full procedure: "scan existing coverage, agree a research rubric, capture every source verbatim before analyzing, write the article incrementally so a crash never loses work, cite every claim, and link it back into the graph" — comprehensive coverage with no material gaps. Not a 4: the action list spans the entire workflow end-to-end, not just several steps with minor gaps.

Completeness

Explicitly answers both what ("Investigate a topic against preserved sources and write a draft-status research article under `research/`...") and when ("Read when asked to research a topic, compare options, synthesize sources, gather evidence, or extend an existing research doc") with concrete trigger phrases. Not a 4: the when-clause enumerates specific triggers rather than a generic "use when working with".

Trigger Term Quality

Good natural-phrase coverage in "Read when asked to research a topic, compare options, synthesize sources, gather evidence, or extend an existing research doc" — five phrasings users would actually say. Not a 5: common synonyms like "look into", "dig into", "survey", or "what are the options for X" are missing; not a 3: the present terms are varied and natural rather than generic keywords.

Distinctiveness Conflict Risk

Clear niche (provisional research articles in a Knowledge Base project) with an explicit boundary — "Does not promote findings to canonical knowledge — that is the sibling `consolidate-notes` skill". Not a 5: the broad trigger "research a topic" could overlap with other research/analysis skills outside the knowledge-base pack; not a 3: the sibling disambiguation and project scoping keep overlap risk minor.