The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several standard security auditing tools to identify vulnerabilities in project dependencies. These include npm audit, pip-audit, dotnet list package, composer audit, and Maven/Gradle dependency checks. All commands are standard diagnostic tools and are used within their intended security context.
[COMMAND_EXECUTION]: The skill performs Git operations to isolate security fixes, such as creating new branches (git switch -c) or creating Git worktrees (git worktree add). These operations include human-in-the-loop checkpoints before execution.
[INDIRECT_PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it ingests and processes data from external sources.
Ingestion points: Outputs from security audit tools (npm audit, pip-audit, etc.) and manual security reports provided by the user.
Boundary markers: The instructions do not define specific delimiters or "ignore instructions" markers when parsing tool outputs into the findings table.
Capability inventory: The skill has the ability to write files (docs/security/, .ways/state.json), execute shell commands (auditors, Git), and propose Pull Requests.
Sanitization: There is no explicit mention of sanitizing or escaping the data parsed from audit tool outputs before presenting them in the Markdown findings report.
[SAFE]: The skill implements a robust "Gate" system (human-in-the-loop), requiring the user to manually approve the scope of findings, the remediation plan, and the final resolution before code changes or Pull Requests are finalized.

fmway-vuln