The Agent Skills Directory

[SAFE]: The skill provides a methodological framework for identifying security gaps in design documents, such as missing authentication models, trust boundary violations, and architectural risks. It does not contain executable code, scripts, or remote dependencies.
[INDIRECT_PROMPT_INJECTION]: The skill processes implementation plans which are external data sources. While this presents an attack surface where a plan could contain instructions to influence the agent, the skill lacks high-risk capabilities like network access, file system writes, or shell execution, limiting the impact of such an injection. The workflow is restricted to generating a security report.
[EXTERNAL_DOWNLOADS]: The skill references local schema files (../../shared/schemas/flags.md and ../../shared/schemas/findings.md) for configuration and formatting. These are relative path references to the local environment and do not involve external network requests or remote downloads.

review-plan