The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell commands including git for repository status and diffing, as well as project-defined validation tools like pnpm, cargo, or ruff. These operations are conducted within the local environment as part of the primary code review and fix workflow.\n- [PROMPT_INJECTION]: The skill analyzes uncommitted code and local files, creating a surface for indirect prompt injection where malicious instructions inside the codebase could attempt to influence the agent's analysis.\n
Ingestion points: Changed files and git diffs are read from the repository in Phase 2 using git diff.\n
Boundary markers: The skill does not employ specific delimiters or warnings to isolate ingested code content from the agent's instructions.\n
Capability inventory: The skill has permissions to write to the filesystem and execute shell commands for linting and fixing as described in Phase 1 and Phase 5.\n
Sanitization: There is no sanitization of code content before it is processed by the analysis agents.\n- [SAFE]: The skill utilizes dynamic context injection (!command) in SKILL.md to display the current git branch and a summary of uncommitted changes. These commands are benign, use hardcoded parameters, and do not incorporate user input.\n- [SAFE]: The workflow incorporates a human-in-the-loop requirement, explicitly halting for user approval before proceeding to apply any automated fixes to the source code.

polish