The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input as search terms for a command-line tool, creating a surface for potential injection.
Ingestion points: User-supplied "query words" are used to populate arguments for the omh memory rejected-recall command (detailed in the Recovery Notes section of SKILL.md).
Boundary markers: While the skill instructs the agent to use "claim boundaries" and "not-evidence boundaries" in its output to distinguish AI-generated guidance from system data, there are no explicit boundary markers or delimiters defined for the user input passed to the tools.
Capability inventory: The skill has access to several omh CLI tools, including omh memory rejected-recall, omh memory project-identity show, and omh runtime record for recording status and retrieving data.
Sanitization: The instructions do not specify any validation, escaping, or filtering for the <query words> provided by the user before they are interpolated into the shell command.

omh-decision-recall