Conciseness

The body is ~37 dense lines with no explanation of concepts Claude already knows and no padding — it does not explain what Terraform, ECS, or Copier are, and the pinned 'copier==9.19.1' is functional within an executable command rather than stale documentation. Every sentence carries an instruction, an exit-code contract, or a guardrail, matching the lean 5 anchor.

Actionability

Three copy-paste-ready commands are given with full argument shapes ('profiles.py resolve <pattern> <runtime> <architecture> <packaging> --region <aws_region>', 'uv run --with copier==9.19.1 python .../render.py <pattern> ./my-service --answers answers.json', 'check_new_project.py ./my-service'), with concrete per-pattern defaults ('python3.13 arm64 zip for apigw-lambda') so the commands run without guesswork. It is not a 4 because the guidance is complete rather than having minor gaps.

Workflow Clarity

The 8-step workflow is clearly sequenced (inspect, pick pattern, check profile, gather answers, render, install harness, check, report) with explicit validation gates and feedback loops: stop on existing project, 'Exit 0 only' profile rule, render refusal semantics (exit 3), and 'Every line must be PASS... UNVERIFIED means a tool or the network was missing: install it and rerun'. This matches the anchor requiring explicit validation steps and error-recovery loops.

Progressive Disclosure

The SKILL.md is a lean overview with two clearly signaled, one-level-deep references ('references/patterns.md lists what exists and when each fits'; 'Read references/runtime-selection.md before proposing a new profile') — both verified to exist with no nested .md chains — and three scripts invoked directly by command. The large patterns/answers tables are correctly split out of the overview, matching the well-structured 5 anchor.

Specificity

The description lists multiple concrete actions ('choosing the pattern, runtime profile and inputs; rendering the tested Copier template; and checking the result') anchored to named technologies (Lambda, ECS, SQS, Step Functions, EC2, Terraform), giving comprehensive coverage of the skill's scope. It is not a 4 because, beyond the mildly soft 'checking the result', there are no material gaps in the action coverage.

Completeness

It explicitly answers both questions: 'Use when starting a new AWS application in Terraform' gives a concrete 'when', and the action enumeration (choosing pattern/profile/inputs, rendering the tested Copier template, checking the result) gives a clear 'what', with an added explicit exclusion ('Not for changing an existing deployed project or for EKS').

Trigger Term Quality

Natural trigger terms are present ('starting a new AWS application in Terraform', 'HTTP API on Lambda', 'ECS web services, SQS workers, Step Functions APIs or EC2 services'), which users would readily say. It falls short of 5 because common synonyms such as 'scaffold', 'new project', 'serverless', or 'infrastructure as code' are missing.

Distinctiveness Conflict Risk

The niche is clear — creating new AWS projects by rendering a tested Copier template — with distinct triggers and an explicit negative boundary ('Not for changing an existing deployed project or for EKS') that minimizes overlap with generic Terraform or AWS-deployment skills. This matches the clear-niche anchor rather than the minor-overlap anchor at 4.