omh-research-brief
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources such as 'customer notes' and external 'page text files' to generate research briefs.
- Ingestion points: External content is processed via 'scripts/sources.py quote' using the '--from ' parameter, as described in SKILL.md.
- Boundary markers: The skill provides explicit instructions to keep observed sources, inference, synthesis, and retrieval gaps separated.
- Capability inventory: The skill utilizes subprocess calls to the 'omh' CLI and 'scripts/sources.py' for citation management and verification (SKILL.md).
- Sanitization: A citation check is implemented using the 'sources.py verify' script to ensure quotes match the retrieved source text.
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using vendor-provided tools and local scripts.
- Evidence: The workflow involves running 'omh runtime record', 'omh research receipts verify', and subcommands of 'scripts/sources.py'.
- [DYNAMIC_EXECUTION]: The skill constructs shell commands at runtime by interpolating variables like source IDs and verbatim quotes into command templates.
- Evidence: Commands such as 'sources.py quote --text ' are assembled dynamically based on findings extracted from research data.
Audit Metadata