The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes external command-line tools, specifically git and the ai-skillet CLI (commands include ai-skillet map and ai-skillet doctor). During implementation tasks, the instructions also authorize the execution of automation scripts (such as install, sync, or publication commands) discovered within the repository's own metadata and documentation.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks as it is designed to ingest and act upon data from external SKILL.md and AGENTS.md files. Maliciously crafted instructions in those files could attempt to influence the agent's behavior, though the skill provides instructions to prioritize mechanical evidence over design inference and to maintain strict evidence boundaries.
[SAFE]: The skill includes positive security controls, such as a defined 'Evidence Boundary' that explicitly prohibits the agent from accessing Git history, transcripts, caches, or unrelated source code. It also respects host authority rules and requires user confirmation for destructive actions like removals or renames.

skill-harmonization