The Agent Skills Directory

[COMMAND_EXECUTION]: The skill mentions executing local project scripts bun run sync and bun run validate. These are standard repository maintenance tools and do not involve untrusted remote execution.- [INDIRECT_PROMPT_INJECTION]: The skill processes local SKILL.md files as input data.
Ingestion points: Reads content from SKILL.md files for validation and formatting.
Boundary markers: Not explicitly defined in the skill instructions.
Capability inventory: Can modify file content and suggest commit messages.
Sanitization: None explicitly defined; the skill performs structural validation (e.g., line counts, list formatting). This is considered a safe surface for its intended development workflow.

hr-content