eval-contract-writerlisted

# Eval Contract Writer Skill ## Intent Every runnable KIT must produce enough executable validation evidence for CLike EVAL and GATE to judge the REQ deterministically. This skill prevents vague HOWTOs, non-runnable LTC files, shallow tests, fake pass claims, and generated code that cannot be evaluated. ## Use when Use this skill for every KIT that generates or modifies source code, tests, CI artifacts, adapters, infrastructure code, frontend components, backend services, model/RAG behavior, MCP/tooling behavior, or runtime integration behavior. ## Do not use when Do not use this skill for pure prose-only documentation changes with no executable validation path. ## Signals Apply this skill when the REQ, PLAN, TARGET_CONTRACT, FILE_REQUIREMENTS, or selected capabilities mention tests, lint, type checks, build checks, smoke checks, security checks, model quality checks, external runners, LTC, HOWTO, evaluation evidence, or gate expectations. ## KIT Generation Rules The KIT must generate validation artifacts as part of the candidate output. Required candidate paths: ```text runs/kit/<REQ-ID>/ci/LTC.json runs/kit/<REQ-ID>/ci/HOWTO.md ``` When dependencies are needed only for the generated candidate, prefer: ```text runs/kit/<REQ-ID>/ci/requirements.txt ``` or the project-native equivalent, if already established. ## Required LTC Behavior `ci/LTC.json` must be machine-readable and should include: - `req_id`; - `lane`; - `runtime_profile`; - `commands[]` or `case