sciomc
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill acts as an orchestrator that ingest untrusted data (research goals and codebase content) to be processed by subagents. This creates a surface for indirect prompt injection where malicious instructions embedded in the analyzed files could influence agent behavior.
- Ingestion points: Processes user-provided
<research goal>and files within the target environment as specified in research stages (e.g., 'Analyze authentication patterns in this codebase'). - Boundary markers: The skill uses structural tags like
[RESEARCH_STAGE],[FINDING], and[EVIDENCE]to delimit output, but does not provide specific instructions to subagents to ignore potential commands embedded within the data being analyzed. - Capability inventory: The skill uses the
Tasktool to invoke subagents (oh-my-claudecode:scientist) and writes session data, findings, and reports to the local filesystem in the.omc/research/directory. - Sanitization: No explicit sanitization or filtering of external content is mentioned before it is passed to the subagents in prompts.
Audit Metadata