tk-learn
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest external data to generate new functional instructions for an agent, which is a potential surface for indirect prompt injection.
- Ingestion points: The skill accepts inputs including conversations, notes, paths, URLs, and repeated workflows (SKILL.md).
- Boundary markers: It implements a strict 'apply gate' checklist and a mandatory 'STOP' checkpoint that halts execution until human approval is granted.
- Capability inventory: The skill possesses the capability to write and rename files on the local file system (SKILL.md, Workflow step 6).
- Sanitization: The instructions require distinguishing 'verified evidence' from 'unverified user claims' and distilling behaviors rather than copying upstream frameworks wholesale (SKILL.md, references/skill-quality.md).
- [COMMAND_EXECUTION]: The skill performs file system modifications to manage skill artifacts. These operations are governed by 'apply authority' gates that require explicit, current-turn approval for a named candidate and target path before any changes are committed.
Audit Metadata