omh-release-cut
Pass
Audited by Gen Agent Trust Hub on Nov 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from commit histories and PR descriptions to automate versioning decisions and generate release plans.
- Ingestion points: SKILL.md identifies 'changes since the last release' and 'observed evidence for published claims' as mandatory inputs.
- Boundary markers: Not present. There are no explicit instructions or delimiters used to encapsulate or ignore embedded instructions within the ingested data.
- Capability inventory: The skill is defined as a 'planner' and explicitly states that it never tags, publishes, or rolls back (OMH prepares, and the host or CI executes), which significantly limits the consequences of an injection.
- Sanitization: No explicit sanitization, filtering, or validation rules for the ingested content are defined in the instructions.
Audit Metadata