The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a custom protocol runtime (_skill_protocol_runtime.py) that executes local Python scripts via subprocess.run to perform tasks like scaffolding files and checking content.\n
Evidence: _skill_protocol_runtime.py contains multiple calls to subprocess.run to execute scripts defined in the manifest.\n- [EXTERNAL_DOWNLOADS]: The check command uses a helper script (learning_files.py) to verify external citations by performing HTTP HEAD or GET requests.\n
Evidence: learning_files.py defines a fetch function using urllib.request to validate up to 50 URLs per execution.\n- [INDIRECT_PROMPT_INJECTION]: The skill maintains learning state across sessions by reading previously generated HTML files from the .learning/ directory. Malicious content in these files could influence the agent's behavior in subsequent sessions.\n
Ingestion points: references/learning-records.md instructs the agent to read study-plan.html and lesson records at the start of a session.\n
Boundary markers: The skill does not explicitly define delimiters for separating the stored state from current instructions.\n
Capability inventory: The agent can execute shell commands (via the protocol wrapper), write to the filesystem, and perform network requests.\n
Sanitization: Input slugs are validated, but content read from stored HTML records is not sanitized for potential prompt injection.\n- [DYNAMIC_EXECUTION]: The protocol runtime uses dynamic imports to load diagnostic and contract modules at runtime.\n
Evidence: _skill_protocol_runtime.py uses importlib.import_module to load either tools.diagnostics or _diagnostic_contract depending on the environment.

tutor