The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script (scripts/check_profile.py) to validate the structure and activation status of user-provided profile files. This script uses standard libraries for regex-based validation and does not perform network operations, dynamic code execution, or unauthorized file system modifications.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied markdown files (person-dna.md), which is a potential surface for indirect prompt injection. The skill robustly mitigates this by requiring a structural validation gate and providing an 'Information Contract' (references/profile-contract.md) that instructs the agent to strictly map data sources, verify permissions, and ignore instructions embedded within the processed data.
[DATA_EXPOSURE]: The skill handles personal professional profiles but implements strict privacy controls. It explicitly prohibits scanning private directories, duplicating sensitive files into the skill's own environment, and exposing unverified or non-public personal information. It also mandates the anonymization of client data to prevent accidental identification.

yetong-topic