Conciseness

Every line is directive and earns its tokens — e.g., "Do not begin review without `VALIDATION: GREEN`, a live PR, and the implementation report" — with zero concept explanations, filler, or padding. The body assumes Claude's competence throughout, matching the 'lean and efficient' anchor rather than anchor 4, which requires trimmable over-explanation that is absent here.

Actionability

The skill supplies verbatim, copy-paste-ready agent prompts with explicit return contracts ("Return `VALIDATION: GREEN`, the resolved absolute plan path, absolute report path, and PR URL; otherwise return the concrete blocker") and concrete gates for each stage. As an instruction-only orchestration skill its guidance is fully executable with placeholders only where caller input is inherently required, matching the top anchor.

Workflow Clarity

Six numbered stages are clearly sequenced with explicit validation checkpoints before every transition (plan path, VALIDATION: GREEN, canonical review report, READY TO MERGE, green CI) and genuine feedback loops: correction rounds feed the complete review back to the implementer, and a PR-caused CI failure returns to correction and a fresh review. Every checkpoint the anchor-5 example demands is present explicitly.

Progressive Disclosure

The body is a well-organized single-file overview that delegates all detail one level deep to clearly signaled sibling skills (/prp-plan, /prp-implement, /prp-review) and keeps only orchestration-level rules inline — an appropriate split with easy navigation. No bundle files exist (references/, scripts/, assets/ are absent), and nothing inlined here belongs in a separate file.

Specificity

"This is an experimental skill. Never use it unless the user explicitly tells you to invoke /prp-deliver" names no capability or concrete action whatsoever; it is purely a gating instruction, matching the 'entirely vague; no concrete actions' anchor. The imperative/second-person phrasing ('Never use it', 'tells you') also incurs the third-person penalty, but the score is already at the floor.

Completeness

The 'when' is answered explicitly (only on direct user instruction), but the 'what' is entirely missing — the description never states what the skill does. This matches the anchor 'only when is present without what' exactly, and cannot reach 3, which requires a clear 'what'.

Trigger Term Quality

The only trigger language is "invoke /prp-deliver" and "the user explicitly tells you" — a single narrow invocation term with none of the natural phrases a user would say when needing this skill (e.g., deliver, ship, take an issue to a merged PR). Falls just above the 'no natural keywords' anchor but well short of good coverage.

Distinctiveness Conflict Risk

The "Never use it unless the user explicitly tells you" gate makes accidental triggering for the wrong skill essentially impossible, giving minimal conflict risk (anchor 4). It does not reach 5 because anchor 5 requires a clear functional niche with distinct triggers, and the description articulates no niche.