The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on subprocess.run across several files (scripts/dry_run.py, scripts/local_smoke.py, and scripts/check.py) to execute CLI commands. These include uv run and the wagents utility, presenting a significant execution surface within the host environment.
[DYNAMIC_EXECUTION]: The script scripts/check.py performs dynamic path resolution to locate its validation toolkit. It is designed to search for files in ../skill-creator/scripts/asset_toolkit, which is a path relative to the parent directory. Accessing and executing logic from computed paths outside the immediate skill directory is a potential risk vector for executing unverified local code.
[REMOTE_CODE_EXECUTION]: The execution of uv run wagents implies the potential for dynamic fetching and installation of dependencies from remote registries. This introduces a dependency on external package state and the integrity of the remote registries.
[INDIRECT_PROMPT_INJECTION]: The skill processes JSON data produced by the wagents tool to validate installation status. While it implements schema validation, the output influences agent summaries.
Ingestion points: scripts/dry_run.py and scripts/local_smoke.py (parsing JSON stdout).
Boundary markers: Absent; the agent is instructed to treat the structured output as factual evidence.
Capability inventory: subprocess.run calls, temporary directory creation, and environment variable modification (HOME override).
Sanitization: Uses strict JSON schema validation, checking for specific keys and data types.

cross-agent-install-smoke