The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it is designed to ingest and process untrusted data from user-specified documents (such as .tex literature reviews) to update project guidelines.
Ingestion points: Files targeted by the guide_path parameter and files within user-provided review directories.
Boundary markers: None identified; the skill directly incorporates extracted insights into markdown files.
Capability inventory: Modification of local markdown files and execution of a structural validation script.
Sanitization: No explicit sanitization of external text is performed beyond structural validation.
[COMMAND_EXECUTION]: The skill executes a bundled Python script (scripts/validate_guide.py) to verify that the mandatory header structure of the guide file remains intact after updates. This is a local, static validation utility.
[DATA_EXFILTRATION]: The skill includes instructions to report technical issues to the vendor's GitHub repository (huangwb8/bensz-bugs) using the gh command-line tool. This involves reading local bug logs and performing network operations to GitHub, which is a well-known and whitelisted service for this purpose.

research-guide-updater