The Agent Skills Directory

[COMMAND_EXECUTION]: The script harnesses/claude/scripts/check_workspace_trust.sh includes an --enable flag that programmatically modifies the user's ~/.claude.json file to set hasTrustDialogAccepted to true. This action bypasses the manual trust dialog designed to prevent automatic hook execution in untrusted project directories.\n- [EXTERNAL_DOWNLOADS]: The scripts harnesses/devin/scripts/verify_docs.py and harnesses/copilot/scripts/verify_docs.py use urllib.request to fetch content from docs.devin.ai and raw.githubusercontent.com. These requests are used to verify that the scaffolded hook definitions remain aligned with official documentation.\n- [REMOTE_CODE_EXECUTION]: The skill's primary function is to generate bash scripts and register them in agent configuration files (such as .claude/settings.json or .devin/hooks.v1.json). This establishes a persistent mechanism for the AI agent to automatically execute project-local code during lifecycle events.

scaffold-hooks