The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes example bash scripts (validate-bash.sh, validate-write.sh) that demonstrate how to intercept and validate shell commands or file operations before they are executed by an agent. These are defensive tools intended to improve security.
[DATA_EXPOSURE]: The documentation and examples provide specific patterns for identifying and protecting sensitive files (e.g., .env, .ssh/, /etc/) from unauthorized access or modification.
[DYNAMIC_EXECUTION]: The skill provides developer utilities such as hook-linter.sh and test-hook.sh which are used to verify the syntax and safety of user-created hook scripts. These tools encourage the use of 'set -euo pipefail' and proper variable quoting to prevent accidental command injection.
[INDIRECT_PROMPT_INJECTION]: The skill explicitly teaches 'Prompt-Based Hooks' as a method for using LLM reasoning to detect malicious intent in data provided to tools, effectively acting as a security layer against indirect injections.

hook-development