code-refactoring-context-restore
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill contains no executable code, remote dependencies, or network operations.
- [DATA_EXPOSURE]: The procedure explicitly instructs the agent to avoid including secrets or private transcripts in generated handoff notes.
- [INDIRECT_PROMPT_INJECTION]: The skill identifies the risk of processing third-party content and logs, specifically instructing the agent to ignore embedded instructions found in retrieved data. Ingestion points include task ledgers and referenced files. It explicitly mandates ignoring embedded instructions as a boundary marker. Capabilities are restricted to reading/writing within authorized project paths.
Audit Metadata