The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow that ingests and processes external data labeled as 'status evidence', 'scope', and 'known risks'. This creates a surface where malicious instructions could potentially be embedded in the data provided to the agent.
Ingestion points: status evidence, scope, time window, and known risks as specified in the Required inputs section of SKILL.md.
Boundary markers: The skill uses explicit output structures (summary, risks, blockers, priorities) and instructions to 'Tie every status claim to observed evidence' to help delimit the processing of external data.
Capability inventory: The skill utilizes shell execution for the omh runtime record tool and local file access to omh-routing/references/skill-common-rail.md.
Sanitization: The instructions mandate the separation of 'observed facts' from 'risks' and 'blockers', and explicitly warn the agent not to 'infer status from missing evidence', providing a logical barrier against deceptive inputs.
[COMMAND_EXECUTION]: The skill references the use of the omh command-line utility (e.g., omh runtime record --skill ops-review) to log activity within the execution environment. This represents standard tool usage within the intended framework.

omh-ops-review