The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a testing utility, scripts/test-hook.sh, which executes local shell scripts provided by the user to verify hook behavior.
[INDIRECT_PROMPT_INJECTION]: The skill's primary purpose is to process data from tool operations, which introduces a surface for indirect prompt injection attacks.
Ingestion points: Data from tool inputs, tool results, and user prompts are processed by hooks as described in SKILL.md and references/patterns.md.
Boundary markers: The documentation recommends variable quoting in shell scripts and provides structured JSON formats, though it lacks specific instructions for LLM-focused delimiters in prompt templates.
Capability inventory: The skill demonstrates capabilities including shell command execution (scripts/test-hook.sh), network requests via curl, and file system operations as shown in references/advanced.md.
Sanitization: Examples such as examples/validate-bash.sh and examples/validate-write.sh provide logic for identifying and blocking path traversal and dangerous command patterns.
[DATA_EXFILTRATION]: The documentation in references/advanced.md provides examples for integrating hooks with external services like Slack webhooks and databases, which involves sending project data to external endpoints.

Hook Development