The Agent Skills Directory

[PERSISTENCE]: The skill's primary function is to install persistent execution hooks. The scripts/install_hooks.py script overwrites the .git/hooks/post-commit file with a launcher and creates .codex/hooks.json and .cursor/hooks.json files. These configurations trigger the execution of hooks/file_size_hint.py and hooks/git_sync_policy.py during normal development workflows.
[COMMAND_EXECUTION]: Several scripts (scripts/install_hooks.py, hooks/file_size_hint.py, hooks/git_sync_policy.py) use subprocess.run to interact with the system's git binary. While these calls are currently limited to repository metadata retrieval (e.g., rev-parse, symbolic-ref), they provide a functional surface for executing shell commands.
[INDIRECT_PROMPT_INJECTION]: The hook scripts analyze external data which could contain malicious instructions designed to influence the agent's behavior. \n
Ingestion points: The scripts in the hooks/ directory parse tool_input and command payloads provided by the IDE/agent environment (e.g., in hooks/file_size_hint.py and hooks/git_sync_policy.py). \n
Boundary markers: Absent. There are no explicit delimiters or instructions to the LLM to ignore content within the analyzed tool inputs. \n
Capability inventory: The scripts can execute shell commands via subprocess.run and write/read files within the .git directory. \n
Sanitization: The hooks/file_size_hint.py script implements path sanitization by resolving target paths and verifying they are relative to the repository root using pathlib.

setup-harness