improve-skill

Pass

Audited by Gen Agent Trust Hub on Mar 2, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The script scripts/extract-session.js reads session log files from the user's home directory. This script accesses chat history for agents like Claude Code, Pi, and Codex.
  • Evidence: The script reads files from paths including ~/.claude/projects/, ~/.pi/agent/sessions/, and ~/.codex/sessions/.
  • [PROMPT_INJECTION]: The skill facilitates a workflow that incorporates untrusted session data into an agent prompt, presenting a risk of indirect prompt injection.
  • Ingestion points: Data is sourced from local agent logs via scripts/extract-session.js.
  • Boundary markers: The prompt template uses <session_transcript> tags to isolate the untrusted content.
  • Capability inventory: The resulting instructions ask the agent to perform file-write operations based on the processed transcript.
  • Sanitization: No content validation or sanitization of the extracted transcripts is implemented.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 2, 2026, 02:43 AM