save-context

Pass

Audited by Gen Agent Trust Hub on Mar 4, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill requests and uses file system tools (Read, Write, Edit) to manage markdown files within the .context/ directory. This is intended functionality to store user preferences, project notes, and profile information.
  • [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface as it stores arbitrary user-provided text into files that are later re-ingested into the agent's context.
  • Ingestion points: Content from the user conversation is saved to files in the .context/ directory, such as profile.md and current-focus.md.
  • Boundary markers: The skill does not define specific delimiters or instructions for the agent to treat the stored data as untrusted or to ignore embedded instructions.
  • Capability inventory: The agent possesses Read, Write, and Edit file system capabilities and can interact with Notion APIs.
  • Sanitization: No sanitization, escaping, or validation of user-provided content is performed before writing to the context files.
  • [DATA_EXFILTRATION]: The skill mentions integration with Notion for task tracking and research pipeline updates. Notion is a well-known service, and there is no evidence of data being sent to untrusted external domains.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 4, 2026, 07:17 PM