The Agent Skills Directory

[SAFE]: The skill defines a defensive development process that prioritizes security and quality through a multi-agent pipeline (Research, Architect, Implement, Test, Critic, Security).
[SAFE]: Includes a dedicated security review stage (Stage 5) that explicitly audits for OWASP Top 10 vulnerabilities, including injection attacks, broken access control, and insecure deserialization.
[SAFE]: Implements a robust 'Critic Isolation Rule' in references/critic-isolation.md to prevent prompt injection and anchoring bias. This includes mandatory extraction procedures to strip narrative rationales from implementation code before it is passed to the reviewer.
[SAFE]: Enforces strict automated quality gates, requiring zero failures in pytest and mypy --strict before any code is allowed to proceed to the review stages.
[SAFE]: Provides automated checks (via grep) to detect and forbid dangerous execution patterns such as eval(), exec(), and shell=True in subprocess calls.
[SAFE]: All external resource references, such as the GitHub repository link, point to the vendor's own infrastructure (bobmatnyc) and show no signs of typosquattng or impersonation.

code-production-process