store-login-session

Pass

Audited by Gen Agent Trust Hub on Oct 19, 2026

Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill and its associated reference file identify sensitive local file paths, including a persistent Chrome profile located at /home/lachlan/.local/share/echomind-oauth-chrome and private project directories at /home/lachlan/Nutstore Files/Share/LazyEdit/. Accessing browser profiles is a sensitive operation as they contain session cookies, history, and potentially stored credentials.
  • [INDIRECT_PROMPT_INJECTION]: The agent interacts with external web content in browser tabs via a CDP adapter (scripts/studio/cdp.py). This creates a surface for indirect prompt injection where instructions embedded in a web page could attempt to manipulate the agent's actions during the release process.
  • Ingestion points: Content from browser tabs accessed through the CDP endpoint at 127.0.0.1:9497.
  • Boundary markers: The skill defines "Recovery boundaries" and requires the agent to verify the destination page and specific app IDs (6814061525) before proceeding.
  • Capability inventory: Local file system access, browser automation via CDP and VNC, and interaction with credential sources.
  • Sanitization: The instructions explicitly direct the agent to enter values without printing them and to ensure that passwords, keys, and cookies remain outside of tool outputs and logs.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 19, 2026, 06:25 PM