logo-creator
Fail
Audited by Gen Agent Trust Hub on Jul 10, 2026
Risk Level: HIGHCREDENTIALS_UNSAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The scripts 'scripts/remove_bg.py' and 'scripts/vectorize.py' attempt to automatically retrieve API keys by searching the user's sensitive shell profile (~/.zshrc) if they are not found in the environment. Searching local system files for credentials is an intrusive and high-risk behavior that targets user secrets.
- Evidence: 'subprocess.run(['grep', 'REMOVE_BG_API_KEY', os.path.expanduser('~/.zshrc')], ...)' in scripts/remove_bg.py.
- Evidence: 'subprocess.run(['grep', 'RECRAFT_API_KEY', os.path.expanduser('~/.zshrc')], ...)' in scripts/vectorize.py.
- [COMMAND_EXECUTION]: The skill uses the 'subprocess.run' module to execute shell commands like 'grep' and 'curl'. While used for the skill's stated purpose, the implementation involves reading from sensitive user files.
- Evidence: Multiple subprocess calls in scripts/remove_bg.py and scripts/vectorize.py.
- [DATA_EXFILTRATION]: The skill transmits API keys and image data to external services (api.remove.bg and external.api.recraft.ai). While these are the intended services, the automated harvesting of keys from local files increases the risk to the user's environment.
- [PROMPT_INJECTION]: The skill processes untrusted user input (brand names and style descriptions) and interpolates it into prompts for image generation and shell command arguments without sanitization.
- Ingestion points: User requirements gathered in Step 1 of SKILL.md.
- Boundary markers: No markers or 'ignore embedded instructions' warnings are present to delimit user data.
- Capability inventory: Command execution (subprocess.run), file system writes, and network operations.
- Sanitization: No filtering or validation of user-provided strings before they are used in commands or prompts.
Recommendations
- AI detected serious security threats
Audit Metadata