The Agent Skills Directory

[SAFE]: The skill implements a robust security guardrail by requiring mandatory user approval through the AskUserQuestion tool before any implementation code is written.
[SAFE]: All suggested command executions (e.g., swift build, swift test, make docs) are standard developer tools appropriate for the project environment.
[SAFE]: Data handling patterns emphasize strong typing, immutability, and strict JSON schema validation, which are effective defenses against injection and data corruption.
[SAFE]: Credential management instructions correctly identify the use of environment variables (ASC_KEY_ID, ASC_ISSUER_ID) for authentication, following industry best practices for CLI security.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user requirements to generate code. This surface is mitigated by the mandatory architecture review phase where the agent must present its plan for user approval before execution. Capability inventory includes file writing and running local build/test commands. Sanitization is addressed through the enforcement of exact JSON snapshot testing.

implement-feature