The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill contains an automated update mechanism that is susceptible to indirect prompt injection.
Ingestion points: The script scripts/fetch-and-transform-hooks-docs.sh downloads external documentation from multiple URLs to temporary local files.
Boundary markers: The transformation prompt used to process the documentation (claude -p "...Convert the human-facing documentation...") lacks boundary markers or instructions to ignore instructions embedded within the fetched data.
Capability inventory: Transformed content is saved directly to the references/ directory. These files are subsequently used as authoritative grounding context by the AI agent when the skill is active.
Sanitization: The skill does not perform validation or sanitization of the fetched external content before passing it to the model for transformation.
[EXTERNAL_DOWNLOADS]: The scripts/fetch-and-transform-hooks-docs.sh script downloads documentation from well-known technology services.
The script fetches configuration and guidelines from official documentation sites for Claude Code (Anthropic), GitHub Copilot, Cursor, Windsurf, and Amp. These sources are recognized as established developer service providers.

hooks-guide