The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external files and git diffs which may contain untrusted content. This presents a surface for indirect prompt injection where malicious instructions embedded in code comments could attempt to influence the agent's behavior. * Ingestion points: Reads files or git diffs provided by the user or sourced from the repository (SKILL.md). * Boundary markers: The skill does not explicitly mention using delimiters or boundary markers when passing content to the subagent (SKILL.md). * Capability inventory: The skill has the ability to modify the filesystem by deleting comments and refactoring code, and executes git commands (SKILL.md). * Sanitization: No specific sanitization or escaping of the ingested code content is described (SKILL.md).
[COMMAND_EXECUTION]: The skill executes local git commands such as merge-base, log -L, and blame to analyze code history and determine the scope of changes. These are standard operations for code auditing tools.

no-comments