paper-to-skill

Pass

Audited by Gen Agent Trust Hub on Apr 18, 2026

Risk Level: SAFENO_CODE
Full Analysis
  • [SAFE]: No malicious patterns, obfuscation, or unauthorized activities were detected. The skill is logically consistent with its stated purpose and references legitimate GitHub resources for issue reporting.
  • [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface as it ingests untrusted external data (PDF papers) to generate new instruction files. This risk is addressed through a structured workflow.
  • Ingestion points: The skill reads user-provided PDF files or text in SKILL.md (Phase 1).
  • Boundary markers: Absent, however, the skill provides a 'Suitability Gate' in Phase 2.5 to filter content and guide the agent's focus.
  • Capability inventory: The agent has the capability to write and save new SKILL.md files to the filesystem in SKILL.md (Phase 4).
  • Sanitization: The skill mandates a 'Self-Verification (Hallucination Check)' in Phase 5, requiring the agent to audit every numerical parameter and claim against the source paper and report results to the user before saving.
  • [NO_CODE]: The skill is composed entirely of natural language instructions, templates, and guides in Markdown format, with no accompanying executable code, scripts, or package dependencies.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 18, 2026, 09:43 AM