The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes instructions to run a bundled shell script (scripts/init-plan.sh) to scaffold a local planning directory. This script performs benign operations such as creating a directory (.plan/), writing markdown templates, and updating the .gitignore file to ensure ephemeral planning state is not committed to the repository. These operations are transparent, well-documented, and consistent with the skill's stated purpose as a workflow tool.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided goals to generate implementation plans. It implements a 'Goal Quality Gate' which forces validation of inputs against specific success criteria (concrete artifacts, evidence, thresholds). This design inherently mitigates risks associated with processing untrusted input by requiring structured validation before content is integrated into the planning output.
[SAFE]: No obfuscation, data exfiltration patterns, credential exposure, or remote code execution risks were detected. The skill follows best practices by explicitly forbidding secrets and machine-specific paths in its specification.

ia-planning