The Agent Skills Directory

[COMMAND_EXECUTION]: The skill configures agent environments to execute a local hook script via the Python interpreter.
Evidence: In scripts/sync.py, the skill defines a command string cmd = f"{sys.executable} {dst}" (where dst is the hook script) and injects it into the configuration files of Claude, Grok, and Cursor.
Evidence: The test suite in tests/test_sync.py uses subprocess.run to execute the sync and hook scripts during automated testing.
[PERSISTENCE]: The skill establishes persistence by registering hooks that run every time the AI agent performs specific actions.
Evidence: scripts/sync.py modifies persistent configuration files including ~/.claude/settings.json, ~/.grok/config.toml, and ~/.cursor/hooks.json to ensure the deny_read_hook.py script is invoked during agent sessions.
[INDIRECT_PROMPT_INJECTION]: The skill processes external configuration data that could be manipulated to influence agent behavior.
Ingestion points: The skill reads a user-defined policy.toml and existing agent configuration files (settings.json, config.toml, cli-config.json) as described in scripts/sync.py and scripts/deny_read_hook.py.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present in the scripts.
Capability inventory: The skill has the capability to write files, create symlinks, modify security settings, and install execution hooks across multiple agent harnesses.
Sanitization: The skill uses tomllib and json for structured data parsing but does not perform higher-level validation of the policy content.
[DYNAMIC_EXECUTION]: The skill dynamically configures execution paths for its internal security hooks at runtime.
Evidence: scripts/sync.py calculates the installation path for the hook script and registers this path within the agent's configuration for execution when tools are used or files are read.

sync-agent-policy