The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/add_case_with_session.py uses subprocess.run() to execute a secondary script (share_session.py) for session uploading. The path to this script is resolved via environment variables (LOV_SHARE_SESSION_SKILL_DIR), hardcoded default locations, or a user-provided argument.
[DYNAMIC_EXECUTION]: scripts/add_case_with_session.py performs dynamic loading of a local helper script (add_case.py) at runtime using importlib.util.module_from_spec. This is used to share validation logic between the main workflow and the standalone mutation script.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted JSON data provided in user-initiated workflows to generate public skill cases, presenting a potential injection surface.
Ingestion points: scripts/add_case.py reads JSON input from standard input or local file paths provided during execution.
Boundary markers: The workflow relies on explicit user confirmation markers rather than technical prompt delimiters.
Capability inventory: The skill has capabilities to write files (cases/cases.json) and execute shell commands.
Sanitization: The skill includes defensive code in scripts/add_case.py that uses regular expressions to detect and block secrets (API keys, Authorization headers, Private Keys) and absolute user filesystem paths from being included in the generated output.
[EXTERNAL_DOWNLOADS]: scripts/verify_public_case.py uses urllib.request.urlopen to perform network operations. These are used to fetch public case registries and verify the existence of artifacts on the vendor's official website and whitelisted code repositories.

lov-skill-add-case