save-session
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The documentation references a design source at
https://github.com/fredoliveira/agents/blob/main/skills/done/SKILL.mdas an attribution for the session recap logic. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the conversation history and tool outputs to generate session recaps. This creates a surface for indirect prompt injection if the source data contains malicious instructions. However, the skill implements several mitigations:
- It mandates a manual preview and user approval step before any file is written.
- It explicitly instructs the agent to exclude credentials, secrets, and unnecessary personal data from the output.
- It follows a strict markdown template which limits the scope of the generated content.
Audit Metadata