The Agent Skills Directory

[PROMPT_INJECTION] (HIGH): Category 8: Indirect Prompt Injection surface. The skill reads arbitrary codebase content and possesses the capability to execute commands and modify files.
Ingestion points: Project files and git history read via Bash, Read, Glob, and Grep tools.
Boundary markers: Absent. The instructions do not define delimiters for codebase content when passed to the subagent.
Capability inventory: Extensive tools including 'Bash' (command execution), 'Write'/'Edit' (file modification), and 'Task' (subagent spawning).
Sanitization: Absent. Untrusted codebase content is analyzed without specific filtering or sanitization, allowing potential embedded instructions to influence the agent or subagent.
[EXTERNAL_DOWNLOADS] (LOW): Uses npx husky init and npx lint-staged. These commands fetch and execute packages from the npm registry at runtime. While these are standard tools, this is an external execution pattern.
[COMMAND_EXECUTION] (LOW): Uses the Bash tool to modify .husky/pre-commit and run gitleaks commands. These actions are aligned with the skill's stated purpose but involve modifying sensitive developer configuration files.

security