setup-artist
Warn
Audited by Gen Agent Trust Hub on Feb 18, 2026
Risk Level: MEDIUMDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- Data Exposure & Exfiltration (MEDIUM): Potential for credential leakage via Git. The skill instructs the agent to create a .env file in the artist workspace for storing API keys and credentials (per references/env-template.md). Step 9 in SKILL.md executes 'git add -A', 'git commit', and 'git push origin main'. If the environment lacks a robust .gitignore or if the .env file is created in a path not covered by existing rules, the automated workflow will stage and push sensitive secrets to the remote repository.
- Indirect Prompt Injection (LOW): Potential for indirect prompt injection surface. The skill ingests external data (Artist Name, RECOUP.md content) and interpolates it into various workspace files. While this is the intended purpose, the ingestion of untrusted data into files that are later read by other agents creates a surface for multi-step injection attacks. Evidence Chain: 1. Ingestion point: RECOUP.md and user input. 2. Boundary markers: Absent. 3. Capability inventory: File write, git push. 4. Sanitization: Absent.
Audit Metadata