skills/rami-maalouf/skills/t3-fork/Gen Agent Trust Hub

t3-fork

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONDATA_EXFILTRATIONDATA_EXFILTRATION
Full Analysis
  • [DATA_EXPOSURE]: The skill accesses sensitive local file paths to read and copy conversation history and application state.
  • Evidence: Accesses ~/.claude/projects/ for transcript files and ~/.t3/userdata/state.sqlite for session metadata.
  • [COMMAND_EXECUTION]: The script executes local system commands and application binaries to identify processes and manage authentication.
  • Evidence: Runs the ps command to detect session IDs from the parent process list.
  • Evidence: Executes the T3 Code application binary (found in /Applications/T3 Code*.app) to issue and revoke authentication tokens.
  • [DATA_EXFILTRATION]: While the skill performs network requests, they are limited to the local T3 Code server instance.
  • Evidence: Uses fetch and WebSocket to communicate with the local server origin defined in ~/.t3/userdata/server-runtime.json.
  • [PRIVILEGE_ESCALATION]: The skill manipulates local database records to control how the T3 Code application imports session data.
  • Evidence: The insertBlockers and removeBlockers functions perform write operations on the provider_session_runtime table in the local SQLite database to prevent duplicate thread creation.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 08:55 PM
Security Audit — agent-trust-hub — t3-fork