Conciseness

The body is lean throughout: steps are compressed bullet instructions ("Hard requirements: goals + non-goals; in-scope/out-of-scope perimeter; top risks + unknowns") with no explanation of concepts Claude already knows and no padding. Every token earns its place, matching anchor 5; anchor 4 would require instances of over-explanation to trim, and there are essentially none.

Actionability

Concrete, specific guidance throughout: named skill invocations ("Invoke breadboarding skill", "Run create-json-prd skill"), exact paths ("docs/04-projects/<lane>/<id>_<slug>/"), and defined artefacts per step. Not anchor 5: several steps rely on external skills whose inputs/outputs are not specified here, and steps 4-5 ignore the ready-made templates that exist in references/, leaving minor execution gaps.

Workflow Clarity

A 15-step numbered sequence with resume logic (step 0 pickup), mandatory checkpoints (step 6 "Oracle bundle (mandatory)", step 7 "mandatory, before PRD creation"), a review-and-revise feedback loop (steps 8-9 review the oracle pass and update dossier files), a Verification checklist, and explicit GO/NO-GO criteria. Clear sequence with explicit validation and feedback loops matches anchor 5.

Progressive Disclosure

The body itself is well-sectioned (Purpose, Inputs, Outputs, Steps, Verification, Go/No-Go), but the bundle contains references/risk-register-template.md and references/spike-section-template.md that are never referenced anywhere — step 4 says only "Write risk-register.md" and step 5 "Record results in spike-investigation.md" without signaling the templates. This matches anchor 3 ("references present but not clearly signaled"); anchor 4 would require the references to be mostly clearly linked.

Specificity

The description names the domain and concrete artefacts — "shape a project from messy inputs into a de-risked, de-scoped shaped packet (brief, breadboard, risks, spikes) ready for wf-plan, with handoff/pickup boundaries" — going beyond the 1-2 actions of anchor 3, but it centers on a single shaping action rather than the multiple distinct actions of anchor 5. Voice is neutral third-person, so no voice penalty applies.

Completeness

Both questions are explicitly answered: the what ("shape a project from messy inputs into a de-risked, de-scoped shaped packet (brief, breadboard, risks, spikes) ready for wf-plan") and the when ("should only be used when the user uses the word workflow and asks to shape a project"). The when is explicit and concrete, matching anchor 5 rather than anchor 4 where the when could be more specific.

Trigger Term Quality

Good natural-term coverage ("workflow", "shape a project", "messy inputs", "de-risked", "de-scoped", "risks, spikes", "handoff/pickup"), sitting noticeably above anchor 3. Not anchor 5: the gate "should only be used when the user uses the word workflow" is a single literal keyword and misses common variations a user would naturally say ("scope this down", "define the perimeter", "triage this idea").

Distinctiveness Conflict Risk

Clear niche within its skill family, ending "ready for wf-plan, with handoff/pickup boundaries to avoid context rot", but the generic word "workflow" as the trigger gate carries minor overlap risk with unrelated workflow-related skills (CI workflows, workflow tooling). Mostly distinct with minor overlap risk matches anchor 4 rather than the minimal-conflict anchor 5.