space-gemini-image

Fail

Audited by Gen Agent Trust Hub on Jul 4, 2026

Risk Level: HIGHCREDENTIALS_UNSAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill instructs the agent to read sensitive authentication data from the local file path config/secrets.md.
  • [DATA_EXFILTRATION]: The skill transmits the API key retrieved from local storage to an external domain api.apicore.ai via a Bearer token. This domain is not associated with the skill author 'SpaceZephyr' nor is it recognized as a well-known trusted service provider.
  • [COMMAND_EXECUTION]: The skill uses the curl utility to perform network operations, which is used to send the API key and user-controlled content to a remote server.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection by processing untrusted external content and user input without safety boundaries.
  • Ingestion points: User-provided image descriptions and external image URLs processed in SKILL.md.
  • Boundary markers: Absent; the skill does not use delimiters or instructions to ignore embedded commands in the user-supplied prompt or URLs.
  • Capability inventory: Sensitive file reading (config/secrets.md) and network transmission (curl).
  • Sanitization: Absent; user input and URL strings are interpolated directly into the JSON payload of a shell command.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Jul 4, 2026, 11:01 AM
Security Audit — agent-trust-hub — space-gemini-image