sd-recon-engine
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external web sources and social media, creating a surface for indirect prompt injection attacks. Evidence:
- Ingestion points: Web retrieval defined in references/source-ladder.md and SKILL.md Step 3.
- Boundary markers: Present. references/recon-evidence.md mandates source tagging, timestamping, and the separation of facts from inferences to mitigate data poisoning.
- Capability inventory: Data retrieval, filtering, and structural formatting for downstream production skills.
- Sanitization: Instructions require verbatim copying of names and explicit marking of unverified or missing information.
- [EXTERNAL_DOWNLOADS]: The skill retrieves research data from specific industry domains including duanjugongcheng.com, shortdramacast.com, verticaldrama.tv, cherrybowl.io, and trysonar.app. These references are integral to the skill's function as an intelligence engine for the short drama sector.
Audit Metadata