The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes 'suggested fixes' from external Markdown files (.findings.md). Maliciously crafted findings could attempt to inject harmful code into the project. \n
Ingestion points: Findings files located in .chalk/reviews/sessions/{session}/. \n
Boundary markers: None (parses markdown tables). \n
Capability inventory: Uses Edit and Write tools to modify source code. \n
Sanitization: Includes explicit instructions to verify fixes and requires user confirmation before any code is applied.\n- [COMMAND_EXECUTION]: The skill is granted Bash access. Although it encourages the use of the Glob tool for file discovery to avoid injection, the underlying capability to execute shell commands remains available.\n- [SAFE]: The skill implements a robust path validation check, ensuring that all target files for fixes are relative paths within the repository root, effectively mitigating directory traversal risks.

fix-findings