The Agent Skills Directory

[SAFE]: The skill implements a rigorous security-first workflow involving architectural decision records, threat modeling, and adversarial testing to ensure production-ready quality.
[SAFE]: The scripts/audit_evidence.py validation tool uses secure parsing for JSON manifests, enforces relative path constraints to prevent traversal attacks, and specifically filters for sensitive credential-like keys to prevent accidental data exposure.
[SAFE]: The workflow enforces a clear separation of duties between research, architecture, and implementation, ensuring that no single identity can both implement and independently certify changes.
[SAFE]: The evidence contract binds approval to specific hexadecimal revisions and SHA-256 hashes, providing a strong cryptographic link between the audited state and the authorized release.

production-upgrade