The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes local shell commands such as ls plans/ for task discovery and directory management. These operations are limited to a project-specific subdirectory and are consistent with the skill's stated purpose of project planning.
[DATA_EXFILTRATION]: No network operations or access to sensitive credential files (e.g., .ssh, .aws, .env) were detected. File access is restricted to plan documentation and project metadata like CLAUDE.md.
[PROMPT_INJECTION]: The instructions establish a behavioral framework for the agent to follow a specific methodology. There are no attempts to bypass safety filters, extract system prompts, or override core agent constraints.
[EXTERNAL_DOWNLOADS]: The skill does not perform any remote downloads or execute scripts from external URLs. It references internal project resources and neighboring skills (e.g., tdd, mutation-testing) for its workflow.
[REMOTE_CODE_EXECUTION]: There is no evidence of dynamic code execution, runtime compilation, or shell piping from untrusted sources.
[SAFE]: The skill implements a robust safety mechanism by requiring explicit human approval before any plan is finalized or any code is committed to the repository.

planning