session-to-skill-extractor

Pass

Audited by Gen Agent Trust Hub on Sep 18, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted session logs from various agent platforms via load_sessions.py. These transcripts are processed by an LLM to identify and articulate procedures. While this presents an indirect prompt injection surface where a transcript could contain malicious instructions, the skill implements a mandatory human review queue (review-queue/) and a manual promotion script (promote.py) which acts as a robust boundary against the automatic installation of malicious or malformed skills.
  • [COMMAND_EXECUTION]: Several test files (e.g., tests/test_filter.py, tests/test_validator.py) utilize subprocess.run() to execute the skill's own Python scripts for integration testing. These executions are limited to the skill's internal logic and do not involve shell injection or execution of untrusted external commands.
  • [EXTERNAL_DOWNLOADS]: The scripts/adapters/devin.py script communicates with the official Devin API (api.devin.ai) using urllib.request.urlopen to retrieve session data when an API key is provided. This is a legitimate vendor resource used as intended for the Devin host adapter and does not represent an exfiltration or remote code execution risk.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 18, 2026, 07:55 AM
Security Audit — agent-trust-hub — session-to-skill-extractor