The Agent Skills Directory

[SAFE]: The skill implements a strictly instructional workflow for converting a finished document into a planning document. It includes robust guidelines for preserving academic integrity (removing execution outcomes, marking gaps as TODOs) and protecting user privacy (stripping personal data such as names, matriculation numbers, and email addresses).
[SAFE]: The execution logic relies on provided Python scripts for deterministic checks of formatting, citations, and reference validity. These scripts (check.py, validate_refs.py) use standard libraries and make authenticated calls to a well-known service (Crossref) for DOI validation.
[SAFE]: The skill implements protective measures against prompt injection by explicitly instructing the agent to treat the source thesis as 'untrusted input' and ignore any directives embedded within it.

proposal-reverse