Conciseness

The body is dense and assumes Claude's competence — it never explains what a Kanban or portfolio is, and the patterns/rules/gates all carry novel information. Minor trimming opportunities exist: the 'How to Run Me — What to Feed Me' first-person framing and the pattern-3 anecdote caveat ('treat it as an anecdote to calibrate against, not a benchmark') are chattier than needed. Mostly efficient with minor over-explanation, so anchor 4 rather than the lean-every-token-earns-its-place anchor 5.

Actionability

Concrete guidance dominates: each pattern has a Smell and a Move ('merge, make one a sub-issue', 're-cut to the riskiest assumption, or graduate it honestly'), plus a scan-approach selection table and three defined output modes. Minor executable gaps remain: how to compute the 'share of capacity' risk-balance split (initiative count vs. effort/spend) is unspecified, and the stale-status sweep procedure is delegated wholesale to the external base skill. Mostly actionable with minor gaps, so 4 rather than fully-executable 5.

Workflow Clarity

Ordering constraints are explicit ('A stale-status sweep comes before any risk-balance claim', 'Run (or assume) base sniff-test reads... first') and the Quality Gates section is a verification checklist including a retry loop ('correct the stages, then re-read the balance'). However, the end-to-end sequence must be assembled from Prerequisite, Scan Approaches, Rules, and Quality Gates rather than one ordered procedure. Clear sequence with most checkpoints present, so 4 rather than the fully-checkpointed 5.

Progressive Disclosure

No bundle files exist and the inline content (the six patterns, rules, output modes, gates) is the core payload that belongs in SKILL.md. External references — '../sniff-test/SKILL.md' and its 'references/', and '../../CREDITS.md' — are one level deep and clearly signaled in a dedicated References section. Sections are well-organized and easy to navigate, matching the clear-overview anchor rather than the minor-organization-gaps level below. (The referenced sibling skill is not present in this review sandbox, but it is a cross-skill dependency, not a buried or nested reference.)

Specificity

The description names five concrete detection capabilities — 'duplicate/shadowing initiatives, WIP clustering, stale-status sweeps, experiment scope creep, and portfolio risk balance (Explore/Discovery vs. Build & Deliver split)' — with comprehensive coverage of what the skill does. It fits the anchor for multiple specific concrete actions with no coverage gaps, so it is not the level below.

Completeness

It explicitly answers both what (the five board-level patterns) and when ('Use when reviewing a whole board or portfolio, prepping a sponsor/ELT portfolio review, or deciding where the portfolio is over- or under-invested') with concrete trigger phrases. This matches the top anchor exactly, not the 'when could be more explicit' level below.

Trigger Term Quality

Natural user phrasing is covered comprehensively, including synonyms and quoted questions: 'reviewing a whole board or portfolio', 'prepping a sponsor/ELT portfolio review', 'deciding where the portfolio is over- or under-invested', 'AI pilots with no clear return', 'where do we start'. No natural term in this domain is plausibly missing, so it is not the level below.

Distinctiveness Conflict Risk

It occupies a clear niche (cross-initiative portfolio patterns) and explicitly routes away three adjacent skills: 'For sniffing a single initiative, use sniff-test', 'use portfolio-to-product-shift-coach', and 'run ai-activity-to-impact first'. Minimal conflict risk, matching the top anchor rather than the minor-overlap level below.