omh-failure-signal-audit
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data, such as source code, agent traces, and runtime reports, to detect swallowed errors. This creates a surface for indirect prompt injection.
- Ingestion points: The skill ingests "code, frontend/browser behavior, agent traces, or runtime reports" as specified in the instructions.
- Boundary markers: The skill explicitly instructs the agent to produce "not-evidence boundaries" and to "separate prepared guidance from observed platform, runtime, connector, file, memory, or delivery evidence."
- Capability inventory: The skill's capabilities are primarily diagnostic and analytical, using the
omhCLI for status recording and generating audit plan artifacts. - Sanitization: The skill relies on logical separation and evidence boundaries rather than automated sanitization to handle untrusted input.
- [COMMAND_EXECUTION]: The skill includes a shell command for the
omhCLI (omh runtime record --skill failure-signal-audit --harness failure-signal-audit --status started) to record the operational status of the workflow. This is a framework-specific tool associated with the 'Oh-My-Hermes' environment and is used for internal status tracking.
Audit Metadata