AI quality: testing strategy, assisted testing, code review, safety, content detection, docs, workflow automation. Topics: ai-assisted-testing, ai-code-review, ai-content-detection, ai-documentation, ai-safety, ai-testing-strategy, ai-workflow-automation, code-review, llm-evaluation.
Router for AI/code quality work. Resolve ONE topic, Read ONLY its playbook
from routes:, then execute. [DOC]
Any request to assess, test, review, evaluate, or harden code or AI output: test plans, AI-assisted/LLM-judge evaluation, human or AI code review, model safety, AI-vs-human content detection, quality docs, or QA automation. [DOC] Not for: building features, prompt authoring, deployment — route elsewhere.
topic (required; infer from intent, ask only if two routes tie),
depth (quick=essentials, deep=exhaustive with verification per step),
plus the artifact under review (code, model output, spec, pipeline). [CONFIG][CÓDIGO] [CONFIG] [DOC] [INFERENCIA] [SUPUESTO]). [DOC]topic (enums are fixed; never invent one). [CÓDIGO]llm-evaluation;
AI writes/runs tests → ai-assisted-testing; AI reviews a diff →
ai-code-review; human/tool review of a diff → code-review. [INFERENCIA]Routing/gate contracts live in assets/ (routing-matrix.json,
quality-rubric.json); see assets/README.md. [CONFIG]
[CÓDIGO] without an
in-repo referent — else downgrade to [SUPUESTO]. [DOC]deep ran verification at each step; quick stayed at essentials.Avoid: loading >1 route; answering from this router without a playbook;
guessing an ambiguous topic; untagged prose; mixed tag families. [SUPUESTO]
On any of these → stop, re-resolve topic via the disambiguation rules,
re-read the one playbook. [INFERENCIA]
fdad39c
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.