huawei-cloud-agentorchard-find-skills
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local Python script
scripts/search_skills.pyto process search queries. This is the primary mechanism for the skill's functionality. - [EXTERNAL_DOWNLOADS]: The search script fetches skill metadata from the official Huawei Cloud AI Gallery API at
devdata.huaweicloud.com. These requests are read-only and target a public endpoint belonging to the skill's vendor. - [INDIRECT_PROMPT_INJECTION]: The skill processes content fetched from an external API, which is a potential surface for indirect prompt injection.
- Ingestion points: Data is fetched from
devdata.huaweicloud.comvia thescripts/search_skills.pyscript. - Boundary markers:
SKILL.mdcontains explicit instructions for the agent to "Output script results directly" without additional explanation or summaries, which acts as a delimiter to prevent the agent from interpreting the data as instructions. - Capability inventory: The skill uses
urllib.requestto fetch data and standard print statements for output; it does not possess capabilities for file writing or privileged execution. - Sanitization: The script parses JSON metadata (titles and descriptions) from the API. The risk is minimized by the agent's constrained output instructions.
Audit Metadata