agent-skill-stack
Pass
Audited by Gen Agent Trust Hub on Jul 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes local utility scripts using python3 and discovery tools via npx. These executions are limited to internal logic and user-directed searches.
- [EXTERNAL_DOWNLOADS]: The skill references external registries including skills.sh, agentskill.sh, and GitHub for skill discovery. These represent standard community resources for agent extensions.
- [DATA_EXFILTRATION]: A included security script (inventory_skills.py) scans for data exfiltration patterns in other skills as a protective measure. The skill does not perform exfiltration itself.
- [PROMPT_INJECTION]: The skill processes external metadata from SKILL.md files. This ingestion surface is mitigated by comprehensive safety guidelines and reference documents which mandate manual review, staged installation, and threat modeling. Ingestion points: The inventory_skills.py and skill_index.py scripts read external skill files; Boundary markers: Clear hard gates and verification protocols are defined in the instructions; Capability inventory: The skill can modify the local filesystem via stage_install.py and create SVG files; Sanitization: Output is escaped in render_stack_card.py and input is parsed with specific regex tokens.
Audit Metadata