uxc-skill-creator
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill includes a validation script (
scripts/validate.sh) and templates for creating validation scripts. These use standard utilities likerg(ripgrep) to check file structure and content against established safety and quality rules. This execution is scoped to repository validation. - [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for ingesting external data, such as official documentation and API endpoints, to generate skill instructions. This is a potential surface for indirect prompt injection if source documentation is malicious; however, the skill mitigates this through the use of strict structural templates and a mandatory validation step to ensure generated instructions adhere to safe patterns.
- Ingestion points:
references/workflow.mddescribes searching external repositories and official documentation for protocol and authentication details. - Boundary markers: The skill uses fixed Markdown templates in
references/templates.mdto structure and wrap ingested information. - Capability inventory: The skill generates text-based instructions and shell scripts for validation.
- Sanitization: Safety is enforced through pattern-matching checks in
scripts/validate.shthat block deprecated or high-risk command patterns.
Audit Metadata