paper-write

Pass

Audited by Gen Agent Trust Hub on Nov 1, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the Bash tool to initialize paper projects, manage file structures, and execute local scripts. It specifically executes a style extraction helper (extract_paper_style.py) and an embedded Python script for bibliography validation.
  • [EXTERNAL_DOWNLOADS]: The skill performs automated network requests to reputable academic services, specifically dblp.org and doi.org, to fetch publisher-verified BibTeX metadata. These operations are essential for preventing citation hallucinations.
  • [DYNAMIC_EXECUTION]: Includes a multi-step Python script embedded within SKILL.md that is executed at runtime to validate bibliography hygiene. It also dynamically resolves the path to a local style-helper script based on the environment configuration.
  • [INDIRECT_PROMPT_INJECTION]: The skill operates by ingesting user-provided research documents (PAPER_PLAN.md, NARRATIVE_REPORT.md) and external style references. While this represents an attack surface given the skill's file-writing and command-execution capabilities, the logic is constrained to the context of academic document generation, and the risk is assessed as minimal.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 1, 2026, 01:42 AM