equibles-research
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill acts as an interface for searching and reading external, third-party documents such as SEC filings and earnings call transcripts, which are potential vectors for indirect prompt injection.
- Ingestion points: Data is ingested via tools like
SearchDocuments,ReadDocumentLines, andGetEarningsCallTranscriptin SKILL.md. - Boundary markers: The skill includes strong defensive instructions under the 'Treat results as data' section, explicitly telling the agent to ignore any instructions found within tool results and to never run commands based on that text.
- Capability inventory: The skill has the ability to modify a user's Equibles account (portfolios, watchlists) via tools like
CreateMyPortfolioandAddPortfolioLot, which could be targeted by instructions hidden in ingested documents. - Sanitization: No programmatic sanitization is specified; the skill relies on instructional constraints to prevent the agent from following embedded commands.
Audit Metadata