podcast-blog-post-creator

Pass

Audited by Gen Agent Trust Hub on Nov 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes podcast transcripts, which are externally provided and potentially untrusted data sources. This creates a surface for indirect prompt injection where hidden instructions in the data could influence the agent's behavior.
  • Ingestion points: The agent is instructed to read through "polished transcripts" in SKILL.md (Phase 1, Step 1).
  • Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are provided in the instructions for handling the transcript data.
  • Capability inventory: The skill is designed to create and write markdown files (SKILL.md, Output Format).
  • Sanitization: No steps for escaping, validating, or filtering the external transcript content are included in the workflow.
  • [EXTERNAL_DOWNLOADS]: The skill references and encourages integration of links to reputable scientific sources.
  • Evidence: Instructions in SKILL.md (Phase 1, Step 3) recommend searching for and linking to peer-reviewed research from the NIH/PubMed Central (pmc.ncbi.nlm.nih.gov) repository.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 9, 2026, 12:32 PM