huawei-cloud-find-skills

Pass

Audited by Gen Agent Trust Hub on Sep 22, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads an executable CLI tool (skill-quality-cli) from a Huawei Cloud Object Storage bucket during its setup process.\n
  • Evidence: scripts/ensure_cli.sh fetches packages from https://obs-skills-repository.obs.cn-north-4.myhuaweicloud.com/skill-quality-cli.\n
  • Security Note: The installation script implements SHA256 integrity verification against pinned hashes to prevent supply chain attacks.\n- [DATA_EXFILTRATION]: The skill automatically collects and sends telemetry data to vendor servers regarding search results and skill installations.\n
  • Evidence: scripts/search-skills.py transmits "exposure impressions" for all search results to https://devdata2.huaweicloud.com/rest/developer/fwdo/rest/developer/servlet/hdskillservice/v1/obs/findcounts/increment.\n
  • Evidence: SKILL.md Step 3.1 mandates recording installation counts to the same vendor endpoint.\n- [COMMAND_EXECUTION]: The skill executes system commands to check for the presence and version of the Huawei Cloud CLI tool (hcloud).\n
  • Evidence: scripts/check-koocli.py runs hcloud version via subprocess.run to provide environment readiness warnings.\n- [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it processes user-provided search keywords and remote skill indices.\n
  • Ingestion points: User keywords passed to scripts/search-skills.py and remote index.json from GitCode.\n
  • Boundary markers: The SKILL.md documentation provides explicit security warnings to the agent regarding shell command injection risks when handling user input.\n
  • Capability inventory: The skill can write files and set executable bits via shell scripts, and execute the hcloud CLI.\n
  • Sanitization: The search-skills.py script includes keyword expansion and generic keyword filtering logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 22, 2026, 09:13 AM