Menu
Use for running automated LLM codebase audits with Gemini 2.5 Pro, ensuring AGENTS.md compliance, API thread safety, and security hardening.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
Use for GoodQ4All repo runtime audits, clean memory starts, Qdrant cleanup, ingestion validation, local fallback/audio repair, operator-console visibility, and agent-facing documentation truth maintenance.
Use to execute Claude Fable 5 prompts on Google Cloud Vertex AI using prompt caching breakpoints to minimize token usage costs by up to 90%.
Use for GoodQ4All repo runtime audits, clean memory starts, Qdrant cleanup, ingestion validation, local fallback/audio repair, operator-console visibility, and agent-facing documentation truth maintenance.
Typer best practices and conventions. Use when working with Typer CLIs. Keeps Typer code clean and up to date with the latest features and patterns, updated with new versions. Write new code or refactor and update old code.
| name | goodq4all-audit |
| description | Use for running automated LLM codebase audits with Gemini 2.5 Pro, ensuring AGENTS.md compliance, API thread safety, and security hardening. |
Use this skill when performing static and LLM-based compliance and security audits on the codebase.
gcloud auth login
gcloud auth application-default login
project-begood):
gcloud config set project project-begood
gcloud config set billing/quota_project project-begood
Run the native LLM-based audit tool using the base python environment:
conda run -n base python scripts/audit_llm.py --project project-begood
--project: GCP Project ID to charge Vertex AI billing.--region: Google Cloud location region (defaults to us-central1).--output: Path to write the generated markdown audit report (defaults to reports/llm_audit_report.md).When auditing the codebase, ensure that:
GOODQ_DATA_ROOT) instead.ValueError) when required directories or environment variables are not supplied._active_tokens in api/routes/ingest.py) with thread-safe synchronization locks (threading.Lock).except: clauses. Always catch specific exceptions (like json.JSONDecodeError or ValueError) and log descriptive warnings to preserve the "fail visible, not loud" protocol."confirm-123" - historical/backdoor removed) exist in the route handlers.