The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs extensive shell command execution to facilitate its scanning functions.
Executes pnpm run update to keep repository dependencies current.
Uses pipx or uvx to install and run the zizmor security scanner for GitHub Actions.
Runs structural validation using node scripts/fleet/check/paths-are-canonical.mts.
Utilizes various search and analysis tools including git, rg, grep, find, oxlint, and tsc across the codebase.
[EXTERNAL_DOWNLOADS]: The skill downloads and installs software from well-known external sources.
Fetches and installs the zizmor tool from its official GitHub repository releases.
Performs dependency updates through pnpm, interacting with public package registries.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface because it processes untrusted codebase data that could contain malicious instructions.
Ingestion points: Untrusted source code, configuration files (e.g., .env, YAML workflows), and git diffs are read from the repository and processed by AI agents in Phase 7.
Boundary markers: No explicit delimiters or instructions are provided to the agents to distinguish between the scanner's instructions and potentially malicious instructions embedded in the scanned files.
Capability inventory: The skill environment allows shell command execution (Bash), file writing (Write), and complex task management (Workflow).
Sanitization: Codebase content is read directly and interpolated into agent prompts without explicit sanitization or escaping to prevent embedded instructions from influencing the agent's behavior.

scanning-quality