The Agent Skills Directory

[SAFE]: The skill's primary functionality is to help users create educational content like study plans and practice questions. It uses a Python script, scripts/build_tutoring_pack.py, which performs standard file writing operations to a local 'outputs' directory. No network exfiltration or sensitive file access was found.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits a data ingestion surface where user-provided learning goals and topics are interpolated into templates. Mandatory Evidence Chain analysis: 1. Ingestion points: The CONFIG dictionary in scripts/build_tutoring_pack.py, which receives strings based on user requests. 2. Boundary markers: No explicit delimiters or 'ignore' instructions are used for the interpolated fields. 3. Capability inventory: The script performs local file writes to outputs/tutoring_pack.json. 4. Sanitization: The script uses standard json.dump for serialization, which handles technical escaping. The surface is standard for the skill's intended use and is not considered a threat.
[METADATA_POISONING]: The skill contains a file named learning-tutoring_audit_result_v1.json which makes self-referential claims about the skill's safety and production readiness. These claims were treated as data for analysis and did not dictate the final security verdict.

learning-tutoring