maintain-workflow-using-unified
Pass
Audited by Gen Agent Trust Hub on Aug 26, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill implements a routing system that determines which skill files to load (e.g.,
skills/<name>/SKILL.md) at runtime. This dynamic computation of paths for loading instruction sets is a security surface that requires validation to ensure only intended local resources are accessed. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon user commands and natural language triggers to shift its internal state and load additional instructions.
- Ingestion points: User-supplied commands such as
/buildor/designand natural language mentions of Unified workflow activation (SKILL.md, Step 0). - Boundary markers: The skill requires an explicit declaration in the format
Using [tier] tier: [skill-name]...before proceeding with implementation, providing a visible checkpoint (SKILL.md, Step 6). - Capability inventory: The skill possesses the ability to read local configuration files (
skills-router.json,skills-index.json) and trigger the loading of secondary skill files via platform-specific tools like the Claude Code Skill tool. - Sanitization: The instructions do not specify a mechanism for sanitizing user-influenced skill names or validating the resulting file paths before they are used for loading.
Audit Metadata