huawei-cloud-mrs-host-fault-diagnose
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The
scripts/lakewatch_api_client.pyutility usessubprocess.runto invoke a local security binary (/usr/local/seccomponent/bin/CryptoAPI) specifically for encrypting and decrypting service account credentials. This is a restricted, non-arbitrary call necessary for the skill's authentication mechanism. - [INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it ingests untrusted data from various sources.
- Ingestion points: Data is collected from MRS cluster logs (
collect_alarm_log_data) and MRS Manager API responses (access_manager_get). - Boundary markers: The skill does not explicitly define delimiters or specific 'ignore instructions' markers when passing collected log content to the agent context.
- Capability inventory: The skill possesses capabilities for shell command execution (via the LakeWatch API client script) and writing to the local filesystem (for token caching).
- Sanitization: The skill implements sanitization by using keyword filters (e.g., 'ERROR', 'Exception') in API requests to narrow the scope of ingested data, though the final content is still processed by the agent.
- [EXTERNAL_DOWNLOADS]: The skill communicates with external LakeWatch service endpoints to retrieve diagnostic information. These communications are directed at official infrastructure associated with the skill's vendor and are authenticated using encrypted tokens.
Audit Metadata