xy-private-ops

Pass

Audited by Gen Agent Trust Hub on Sep 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill includes a local script scripts/atoms-search.py which is used to search an atomic knowledge base. The agent is instructed to execute this script via a shell command (python3 ...) to retrieve operational guidelines and case studies.
  • [EXTERNAL_DOWNLOADS]: The helper script atoms-search.py performs network operations to https://api.xyskill.xyz/v1/atoms/search. This is a vendor-owned API used to fetch diagnostic data. If the API is unreachable, the script fails over to a local atoms.jsonl file.
  • [INDIRECT_PROMPT_INJECTION]: The skill contains a vulnerability surface where user-supplied keywords are interpolated into a shell command execution.
  • Ingestion points: User input provided during the Phase 1 and Phase 2 diagnostics (e.g., business problems,朋友圈 keywords).
  • Boundary markers: Absent. The instruction does not include specific delimiters or warnings for the shell execution context.
  • Capability inventory: The skill possesses shell command execution capabilities (python3 scripts/atoms-search.py) and network connectivity via the Python environment.
  • Sanitization: The skill relies on simple double-quoting of the user variable ("<关键词>"), which provides insufficient protection against shell metacharacters or complex command injection payloads.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 18, 2026, 05:21 AM
Security Audit — agent-trust-hub — xy-private-ops