fullstack-dev-skills
Warn
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: MEDIUMMETADATA_POISONINGEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [METADATA_POISONING]: The skill's manifest in
SKILL.mdclaims the author is 'Trae AI', which is inconsistent with the platform metadata and repository source ('wuxiaolong123121'). This impersonation of a known AI platform is misleading and appears intended to gain unearned trust. - [EXTERNAL_DOWNLOADS]: The skill's installation process fetches content from a personal GitHub repository ('wuxiaolong123121') that is not a verified or trusted vendor.
- [INDIRECT_PROMPT_INJECTION]: The skill is built to handle full-stack development tasks that involve processing external codebases, documentation, and reference files. It lacks the necessary boundary markers or safety instructions to prevent the agent from executing malicious instructions embedded in that data.
- Ingestion points: Ingests external code, repository structures, and markdown reference files during the development workflow.
- Boundary markers: None present in the main instruction set to delimit external content or warn against embedded instructions.
- Capability inventory: The skill scope includes high-privilege activities such as DevOps (Docker, Kubernetes), CI/CD, and 'Computer Use' automation.
- Sanitization: No logic or instructions are provided to sanitize or escape external content before it is processed by the agent.
Audit Metadata