skills/aliyun/alibabacloud-aiops-skills/alibabacloud-ecs-linux-os-troubleshooting/Gen Agent Trust Hub
alibabacloud-ecs-linux-os-troubleshooting
Pass
Audited by Gen Agent Trust Hub on Aug 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSPROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The agent uses the aliyun CLI
run-commandutility to execute shell scripts on target and rescue ECS instances. This allows for data collection and system manipulation across the environment as documented in SKILL.md and references/aliyun-cli-cheatsheet.md. - [CREDENTIALS_UNSAFE]: The offline recovery process described in references/utils/guestos-pe-prep.md includes instructions for injecting new passwords or SSH keys into an instance's system disk during reattachment. While this is a platform-level requirement for disk reattachment, it involves overwriting existing security credentials.
- [REMOTE_CODE_EXECUTION]: In references/utils/tracepoint-perf-tools.md, the skill fetches performance tracing scripts from an external GitHub repository (github.com/brendangregg/perf-tools) for kernel investigation.
- [PROMPT_INJECTION]: The skill exposes an attack surface for indirect prompt injection through the processing of untrusted system data. Ingestion points include serial console logs (references/utils/guestos-console-log.md), system logs, process metadata (references/utils/guestos-malware-mining.md), and content from external URLs. Boundary markers are partially present in the form of validation guidelines in SKILL.md. Capability inventory includes remote command execution and disk management. Sanitization is addressed by instructions to validate user inputs and use JSON serialization for CLI parameters.
- [DYNAMIC_EXECUTION]: A complex shell script is dynamically assembled in references/utils/guestos-pe-prep.md using string templates to automate the identification and mounting of disk partitions on rescue instances.
Audit Metadata