The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and interpret user prompts regarding tool requirements for workflows, creating a surface for potential indirect injection.
Ingestion points: Processes user requests and target context provided during the readiness check phase (SKILL.md).
Boundary markers: Explicitly mandates the use of "not-evidence boundaries" and "prepared-vs-observed boundaries" to ensure the agent distinguishes between its own instructions and actual system state.
Capability inventory: Involves diagnostic capabilities such as command -v, tool_search, and hermes mcp list to verify tool presence.
Sanitization: Focuses on inventorying and reporting gaps rather than executing untrusted strings; instructions emphasize that readiness guidance is not execution evidence.
[COMMAND_EXECUTION]: The skill utilizes standard diagnostic shell commands (command -v) and the vendor's own CLI tool (omh) to perform environment checks. These operations are scoped to verifying the existence of dependencies and recording workflow status, which is consistent with its stated purpose as a readiness tracker.
[CREDENTIALS_UNSAFE]: The skill demonstrates safe handling of sensitive information by explicitly instructing the agent to list the "env-var NAME only" for credentials and to flag any credentials that are broader than the task requires as "unsafe."

omh-toolbelt-readiness