The Agent Skills Directory

[COMMAND_EXECUTION]: The skill implements logic via bash scripts (enforce-toolchain.sh and session-env.sh) that are executed by the agent environment as hooks. These scripts analyze commands and manage environment state.
[SAFE]: The script enforce-toolchain.sh functions as a security guardrail by intercepting shell commands before execution. It blocks potentially destructive operations such as broad rm -rf commands, git push --force, and git reset --hard unless they target a whitelist of safe directories (e.g., node_modules, .next, dist).
[SAFE]: The session-env.sh script performs routine maintenance, such as cleaning up its own temporary session directories in /tmp/hook-session-* and establishing environment variables for the current session. The file operations are scoped to a specific prefix and directory.
[SAFE]: No network operations, data exfiltration patterns, or obfuscation techniques were identified. The skill operates entirely within the local project context using standard toolchain components (git, bun, jq).

setup-toolchain