The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a project-local CLI tool located at '.claude/scripts/code_graph' to perform graph-assisted investigations of the codebase. This is a standard architectural pattern for development environments and does not involve remote execution of untrusted code.\n- [SAFE]: The framework includes explicit 'Verification Gates' that require the agent to provide fresh evidence (e.g., test outputs or build logs) before claiming a task is complete, preventing false success reports.\n- [SAFE]: The 'Receiving Feedback' and 'Rationalization Prevention' sections guide the agent to evaluate suggestions with technical rigor and avoid common evasions, ensuring high-quality outputs and preventing blind implementation of potentially incorrect external advice.\n- [SAFE]: The skill defines clear architecture boundary checks using 'docs/project-config.json', promoting best practices in dependency management and structural integrity.

code-review