Conciseness

Every line carries operational weight — 'Probes read data; they don't free-associate', 'count name frequency — the 80/20 read', 'An improvement we can't revert or falsify is a mandate, not a bet' — with zero explanation of concepts Claude already knows. Even the aphoristic lines encode rules compactly rather than padding.

Actionability

Concrete, executable guidance throughout: named inputs ('board.md, flow-log.csv, recent reviews/, learnings/, topology.md, and the cards' dependencies: fields'), named outputs ('improvements/<slug>/improvement.md', 'reviews/YYYY-MM-DD-improve-probe.md'), and specified table columns ('% work localized / change size / risk / future-proofness'). Not a 5 because the actual probe definitions and the improvement template live in external files not shown here, and step 5's 'compose your spec-writing skills' is a high-level hint rather than a specific instruction.

Workflow Clarity

A clearly sequenced 6-step workflow (load library → gather data → run probes → capture bets → right-size ceremony → report) with embedded checkpoints in step 3 ('signal present? cite the data. No signal? one "checked, quiet" line') and an explicit Quality Gates checklist closing the loop ('Every probe run is accounted for', 'Zero direct edits to pdlc.md / topology.md / tiers in this invocation'). Not a 4 because validation is explicit and checklist-form, not merely implicit.

Progressive Disclosure

The body is a concise overview that keeps probe definitions external ('skills/portfolio-pdlc/references/probe-library.md'), referenced one level deep and clearly signaled both in workflow step 1 and in a dedicated References section with a one-line description of each target. Not a 4 because there is no inline content that belongs in a separate file and no organization gaps — navigation is trivial.

Specificity

Names the domain and several concrete actions — 'Probe the portfolio operating system itself', enumerating both probe families ('flow boundaries, altitude, WIP, discovery discipline, measure-and-learn'; 'dependency concentration, platform extraction, team-around-outcome options') and the capture mechanism ('spec-driven bets in the improvements/ backlog'). The core actions reduce to probe-and-capture, leaving minor gaps in coverage rather than the comprehensive multi-action list of a 5.

Completeness

Explicitly answers both: what ('probe the portfolio operating system... and capture improvement ideas as spec-driven bets in the improvements/ backlog instead of implementing them on the spot') and when ('Use when the board is healthy enough to earn self-improvement, after recurring findings, or on an improvement cadence') with three concrete trigger conditions. Not the level below because the 'when' clause is explicit and specific rather than vague or merely implied.

Trigger Term Quality

Good natural keyword coverage for this domain — 'portfolio', 'PDLC', 'process', 'WIP', 'topology', 'board', 'improvement', 'backlog', 'self-improvement', 'improvement cadence' — with focus tags as dispatch terms. A few natural variations users might say ('operating model', 'retrospective', 'continuous improvement') are missing, and family-internal jargon like 'Loops 3+4' is not a phrase a user would naturally utter.

Distinctiveness Conflict Risk

Occupies a clear niche — self-improvement of the portfolio PDLC operating system, explicitly positioned as 'Loops 3+4 of the portfolio-pdlc family' with distinct focus tags (focus:process, focus:topology, focus:auto). Triggers ('board healthy', 'improvement cadence') are specific to this skill and would not fire related skills; minimal conflict risk.