Conciseness

The ~40-line body is extremely lean: every rule is a single terse directive ("Do not improvise beyond scope", "Lightweight = small clear tasks") with zero padding and zero explanation of concepts Claude already knows. Every token earns its place, matching the 'lean and efficient' anchor.

Actionability

Most rules are executable directives — "Write to unique file path defined by orchestrator", "If instructions are ambiguous, STOP and ask orchestrator before executing", and a concrete enumerated return format ("concise results, summary, side effects, anomalies, discoveries, contract changes, deviations, inconsistencies, and insights"). Minor gaps remain: rules like "Keep standard agent tools available as required" and "Initialize required skills on start" are vague, and there is no example of what a compliant return/output actually looks like.

Workflow Clarity

The body is organized as a lifecycle (Identity → Input contract → Output contract → Behavior) with explicit checkpoints: a pre-execution gate ("If instructions are ambiguous, STOP and ask orchestrator before executing") and an execution gate ("MUST STOP and EXPLAIN if cannot execute as requested or off-plan"). It is a behavioral contract rather than a numbered procedure, so the sequence is implicit rather than spelled out step-by-step, keeping it below the anchor-5 pattern.

Progressive Disclosure

The skill is under 50 lines with no need for external references (no references/, scripts/, or assets/ bundle exists), and the body is well-organized into clearly signaled sections (process with Identity/Input/Output/Behavior subsections, plus a pitfalls section). Per the rubric's simple-skill guidance, this merits a 5 with well-organized sections alone.

Specificity

The description lists four named deliverables ("Defines your input contract, output contract, behavior boundaries, and escalation protocol"), but these are abstract categories rather than concrete actions a reader can picture executing — closer to "Processes PDF files and extracts content" than to "Extracts text, fills forms, converts pages". It also uses second person throughout ("you ARE a subagent", "your input contract"), which the guidelines penalize by reducing the specificity score by 1 from an already borderline 3/4 base.

Completeness

Both 'what' ("Defines your input contract, output contract, behavior boundaries, and escalation protocol") and 'when' ("MUST activate when you ARE a subagent") are explicitly answered, with three concrete trigger conditions in the when-clause. This matches the anchor-5 pattern of explicit what-and-when with concrete trigger phrases, not the anchor-4 pattern where the when is generic.

Trigger Term Quality

Natural trigger phrases are well covered: "you were spawned by an orchestrator", "you received a delegated task", "you are executing within a subagent context" — terms an orchestrator or user would naturally say. A few natural synonyms (e.g., 'worker', 'delegation', 'fan-out task') are missing, so it falls short of the comprehensive anchor 5.

Distinctiveness Conflict Risk

The trigger condition — being a spawned subagent inside an orchestration — is a clear niche with distinct triggers and minimal overlap risk with ordinary user-facing skills. Even other orchestration skills would not fire on 'you ARE a subagent' phrasing.