Conciseness

At 544 lines the body restates the same rules repeatedly: the ultra opt-in / 'explicit regular restores the save prompt' rule appears in Mode Selection, Artifact Ownership context, the plan-exists section, and Guardrails; the verbatim-marker preservation list is enumerated twice (in the Step 0 language contract and the regular-mode save section) and again in ULTRA-RESEARCH-FORMAT.md; the language/technical-terms policy consumes dozens of lines. This matches anchor 2 ('noticeably verbose; several unnecessary explanations or padded sections') rather than 3 — the redundancy is systemic, not occasional.

Actionability

Guidance is largely executable: exact config keys and read order in Step 0, an exact three-option save prompt, a complete research-file skeleton with literal markers, a concrete `mkdir -p "$(dirname ...)"` command, and a numbered 6-step ultra procedure. Minor gaps keep it from 5: the Research Coherence Gate criteria are partly abstract ('understandable without the conversation'), subagent delegation is hedged ('when supported'), and the 'Bundle Integrity Gate' is invoked without local definition.

Workflow Clarity

Sequences are explicit and include real validation checkpoints with feedback loops: the Research Coherence Gate lists four numbered verification criteria, mandates 'Correct or qualify mismatches ... then re-run', and gates session-append on success; the regular save procedure is step-by-step with an option-driven prompt. It falls short of anchor 5 because execution order is scattered across topic-organized sections (config, ownership, mode selection, gates, save) with no single flow summary, and one referenced gate is undefined in this file.

Progressive Disclosure

The bundle structure is solid: three real, well-scoped reference files (ULTRA-RESEARCH-FORMAT.md, PROJECT-HYPOTHESES.md, RETRY-HYPOTHESIS.md), each one level deep and explicitly condition-gated ('In ultra mode, read references/ULTRA-RESEARCH-FORMAT.md completely before choosing a slug'; 'Only when the user asks to ... track a project hypothesis'). It is not 5 because substantial ultra-contract detail (marker names, status values, compatibility headings, inclusion rationale) is inlined in SKILL.md and duplicated in ULTRA-RESEARCH-FORMAT.md rather than living solely in the reference.

Specificity

Names the domain ("investigate ideas, systems, problems, and requirements before planning") and a concrete artifact outcome ("research bundle with an index"), but the action verbs are generic and "durable adaptive research bundle" is jargon-heavy; coverage of the skill's actual capabilities (codebase exploration, option comparison, diagrams) is partial. It lists fewer, more abstract actions than anchor 4's 'several specific actions', matching anchor 3's '1-2 concrete actions, not comprehensive'.

Completeness

The 'what' is clear ("Enter explore mode to investigate ideas, systems, problems, and requirements before planning"), but the only explicit 'when the user wants...' clause is scoped to selecting ultra mode within the skill, not to invoking the skill itself; the skill-level 'when' is only weakly implied by 'before planning'. Per the guideline capping weak/implied triggers, this matches anchor 3 ('clear what but when is missing or only weakly implied') rather than anchor 4, whose when-clause addresses skill usage.

Trigger Term Quality

"explore", "investigate", "ideas", "problems", "requirements", and "research" are terms a user might naturally say, but "ultra mode", "config", and "C4, ADR, or dependency artifacts" are internal jargon, and common natural variations (brainstorm, discovery, think through, compare options, spike) are absent. This matches anchor 3 ('some relevant keywords but missing common variations or synonyms') rather than anchor 4's 'good keyword coverage'.

Distinctiveness Conflict Risk

"Explore ... before planning" plus the research-bundle/C4/ADR artifact framing carves a clear niche distinct from general-purpose skills, with only minor overlap risk against generic brainstorming or research skills — matching anchor 4 ('mostly distinct; minor overlap risk'). Not 5 because the description omits the /aif-plan toolchain context that would make the niche unambiguous, and 'explore ideas and problems' alone could collide with a generic ideation skill.