The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill workflow requires the agent to read and execute instructions from documentation files like AGENTS.md and CLAUDE.md within the audited repository to identify valid test runners and commands. This introduces a surface where malicious instructions placed in those project files could lead to unauthorized actions.
Ingestion points: The agent is instructed in SKILL.md to use test commands documented in the repository's AGENTS.md, CLAUDE.md, or package scripts.
Boundary markers: No explicit constraints or validation steps are provided to ensure the commands found in project documentation are safe or restricted.
Capability inventory: The agent is authorized to execute git commands and project-specific test runners.
Sanitization: There is no mention of filtering or sanitizing the content retrieved from external project files before execution.
[COMMAND_EXECUTION]: The skill guides the agent to perform various command-line operations, including running test runners, project scripts, and git utilities (git diff --check, git diff --numstat).
[NO_CODE]: The skill consists entirely of markdown documentation, configuration YAML, and static assets, with no included executable scripts or compiled binaries.

test-audit