ulw-plan
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which can contain malicious instructions.
- Ingestion points: User requirements, codebase facts, and web/source research data are ingested into the planning context to generate alternatives and risk reviews.
- Boundary markers: The instructions do not specify explicit structural delimiters or 'ignore' warnings for all ingested data, though it requires output to be recorded in specific Markdown artifacts.
- Capability inventory: The skill has the ability to execute shell commands (via the
omhCLI) and write files to the local filesystem (under.omh/plans/). - Sanitization: No specific escaping or validation logic is mentioned for user-supplied task descriptions or codebase data interpolated into commands.
- [COMMAND_EXECUTION]: The skill uses a vendor-specific CLI tool (
omh) to record runtime evidence, manage planning states, and finalize plan acceptance. - Evidence:
omh runtime record --skill ralplan --harness planning --status started,omh hermes plan --record, andomh hermes plan-accept <path>.
Audit Metadata