skillify
Pass
Audited by Gen Agent Trust Hub on May 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill directs the agent to write new markdown files to the local file system, specifically targeting
${CLAUDE_CONFIG_DIR:-~/.claude}/skills/omc-learned/or.omc/skills/. This creates persistent agent capabilities that will be active in future sessions. - [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through its ingestion of session history.
- Ingestion points: The skill processes the current session's conversation and actions as the source for the new skill draft (SKILL.md).
- Boundary markers: There are no boundary markers or instructions to treat session data as untrusted while extracting the workflow.
- Capability inventory: The skill allows the writing of new skill files to directories used by the agent for loading tools and instructions.
- Sanitization: The skill lacks technical sanitization or validation steps to ensure that malicious instructions in the session history are not converted into permanent skill logic.
Audit Metadata