chain-method
Pass
Audited by Gen Agent Trust Hub on Aug 25, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill operates as an educational tool for memory techniques, using self-contained logic and local reference data. It does not perform external network calls or download untrusted code.\n- [COMMAND_EXECUTION]: The instructions direct the agent to provide a 'vidbyte retain' command template to the user. Security is ensured by prohibiting the agent from auto-executing the command and requiring it to be safely quoted, maintaining user control.\n- [DATA_EXFILTRATION]: Mnemonic chains and drill results are saved to local markdown files within the skill's workspace. These operations do not involve exfiltration of sensitive information or unauthorized file access.\n- [PROMPT_INJECTION]: The skill ingests user-provided lists for memorization. The potential for indirect injection is managed through a secure evidence chain:\n
- Ingestion points: User input via the /chain-method command arguments.\n
- Boundary markers: The skill uses markdown formatting and sequential phases to isolate instructions from user data.\n
- Capability inventory: Capabilities are restricted to local file writing for task-specific outputs.\n
- Sanitization: Input is normalized and concrete symbols are agreed upon before mnemonic links are generated.
Audit Metadata