The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various system commands including git, bunx, and sha256sum to inspect and modify the environment.
[DYNAMIC_EXECUTION]: The maintenance script scripts/update-skills.sh is designed to execute an external bootstrap script located within the user's dotfiles directory (~/dotfiles/.chezmoiscripts/run_after_install-global-skills.sh).
[DATA_EXFILTRATION]: The skill includes instructions to perform git push operations on both the project repository and the user's personal dotfiles repository. This capability could be exploited to exfiltrate sensitive data if an attacker can influence which files are staged for commitment.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The agent is instructed to read and interpret the output of git status and dots status in SKILL.md to determine the state of the repositories.
Boundary markers: No delimiters or instructions are used to distinguish between legitimate system status and potentially malicious content embedded in the environment (e.g., in branch names or filenames).
Capability inventory: The skill allows the agent to execute shell commands, run local scripts, and push code to remote servers, providing a significant impact path for successful injections.
Sanitization: Data retrieved from external command outputs is processed without sanitization or validation.

skills-maintenance