paper-workflow
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, untrusted data in the form of manuscript drafts, sections, and PDF content provided by users.
- Ingestion points: Inputs include individual sentences, paragraphs, full sections, or entire manuscript drafts provided in the agent's context (SKILL.md).
- Boundary markers: The instructions explicitly mention carrying forward "named protected passages" and respecting "explicitly frozen decisions," which serves as a logical boundary for the AI (SKILL.md).
- Capability inventory: The skill has the capability to revise text and suggest edits to manuscript files.
- Sanitization: There are no explicit instructions for sanitizing or escaping the content of the manuscripts, but the skill includes strong instructions against fabricating data or results, which mitigates standard injection risks.
- [REMOTE_CODE_EXECUTION]: The documentation mentions that an optional specialist skill (
reference-audit-guide) "ships runnable verification scripts." However, this specific skill (paper-workflow) does not contain or execute those scripts itself, and it explicitly states that its own data-gathering script (scripts/section_corpus.py) is "not part of the installed skill" (references/section-evidence.md).
Audit Metadata