The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides operational scripts such as check_workspace_trust.sh and check_hooks_feature.py that directly modify global configuration files of the host AI agents (e.g., ~/.claude.json and ~/.codex/config.toml). These scripts can programmatically set flags like hasTrustDialogAccepted=true, which bypasses the built-in interactive security dialogs that agents use to confirm if a user trusts a specific repository before executing hooks or tools.
[COMMAND_EXECUTION]: The primary function of the skill is to generate and install numerous shell scripts into a project's hooks/ directory and wire them into the agent's runtime environment (e.g., through .claude/settings.json or .codex/hooks.json). This creates a persistent execution mechanism where the AI agent will automatically run project-local scripts during various lifecycle events, such as before tool execution (PreToolUse) or when the agent stops its turn (Stop).
[EXTERNAL_DOWNLOADS]: The OpenCode harness component configures the agent to load the opencode-froggy plugin, which is a third-party extension. While the sources are identified as GitHub and npm (well-known services), this configuration introduces an external dependency into the agent's operational environment that is not maintained by the primary agent vendor.

scaffold-hooks