multi-source-synthesis
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill operates as a 'reduce' step that merges findings from multiple individual summaries, creating an attack surface for indirect prompt injection where malicious content within those sources could influence the synthesis process or the agent's behavior.
- Ingestion points: The skill ingests summaries generated by subagents or delivered via teammate messages (SKILL.md).
- Boundary markers: The instructions do not prescribe the use of specific delimiters or 'ignore' instructions to isolate source content during the synthesis workflow.
- Capability inventory: The synthesis workflow involves high-level reasoning tasks (deduplication, clustering, and conflict surfacing) that are vulnerable to manipulation by maliciously crafted input data.
- Sanitization: No explicit sanitization, validation, or escaping routines are specified for the input text before it is integrated into the final output.
Audit Metadata