Conciseness

The body explains nothing Claude already knows and defers all domain detail to references; nearly every token is directive. It falls short of anchor 5 because several sentences are convoluted or redundant ('Labelling an unsupported operational default as an assumption does not authorize using it when that authorization is absent') and could be tightened without loss, but it is well above anchor 3's unnecessary-explanation level.

Actionability

Concrete, executable directives dominate: which file to read before which action ('Read references/pn-construction.md and references/checks.md before beginning'), named tools ('mutate_workpiece', mounted Petrinaut tools), a three-way disposition rule (Changed/Already represented/Blocked), and explicit authorization conditions. Gaps keep it from anchor 5: 'Orient' and 'Elicit or revise' remain fairly abstract, and no concrete example of a workpiece entry or net fragment appears anywhere in the body.

Workflow Clarity

The lifecycle is explicitly sequenced (Before substantive work → Orient → Elicit or revise → Maintain the workpiece → Construct → Check and deliver) with the interleaving model stated up front, explicit validation checkpoints ('run references/checks.md' after each net change, 'Apply references/checks.md whenever construction is prepared or attempted'), a feedback loop (Blocked → ask the smallest resolving question or record the gap), and the disposition rules acting as a per-answer checklist. Matches anchor 5; anchor 4 would require missing checkpoints, which are not missing.

Progressive Disclosure

The body is a genuine overview that signals each one-level-deep reference at its exact point of need (profile.md before elicitation, pn-construction.md and checks.md before construction, experiment-configuration.md on decision statements) and explicitly forbids recursive following; the four references exist and do not chain further. It misses anchor 5 because 'templates/workpiece.md' is referenced twice but absent from the bundle, and the resource-discipline instruction (paths only via the activation briefing, relative names as labels) makes navigation less direct than a standard well-signaled reference layout.

Specificity

The description lists several concrete actions — 'Elicit or revise an operational process model', 'maintain its recoverable workpiece', 'construct a checked SDCPN' — with minor coverage gaps (e.g., checking/analysis is only implied). It sits between anchor 3 (1-2 concrete actions) and anchor 5 (comprehensive action list), closer to 4; some actions like 'maintain its recoverable workpiece' are abstract jargon rather than a fully concrete operation.

Completeness

Both parts are explicit: 'what' ('Elicit or revise... maintain its recoverable workpiece, and construct a checked SDCPN') and 'when' ('Use for a process-modelling interview, Petri net, or analysis or revision of either artifact'). Not anchor 5 because the 'when' clause is terse and jargon-bound — the conditional 'when Petrinaut capabilities are available' is a capability gate rather than a user trigger, and trigger phrases could be more specific.

Trigger Term Quality

Natural trigger terms are present — 'process-modelling interview', 'Petri net', 'analysis or revision' — and a user needing this skill would plausibly say 'Petri net'. It falls short of anchor 5's comprehensive synonym coverage (missing variations like 'process model', 'workflow model', or tool-specific names alongside the jargon 'SDCPN'/'Petrinaut').

Distinctiveness Conflict Risk

The SDCPN/Petri-net process-modelling niche is clearly staked out with distinct triggers ('process-modelling interview', 'Petri net'); it would not plausibly activate the wrong skill. Matches anchor 5's clear-niche example, and is not the slightly-overlapping anchor 4.