skills/erphq/skills/dark-mode/Gen Agent Trust Hub

dark-mode

Pass

Audited by Gen Agent Trust Hub on Jul 6, 2026

Risk Level: SAFECREDENTIALS_UNSAFEDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill instructs the agent to retrieve authentication tokens from a sensitive local file path at ~/.build-host/credentials.json.
  • [DATA_EXFILTRATION]: The skill enables the transmission of project data to an external API endpoint POST /api/projects/:uuid/deploy during the deployment phase.
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection because it requires the agent to read and process existing UI code which could contain adversarial instructions.
  • Ingestion points: The agent parses existing project UI and design constraints (SKILL.md).
  • Boundary markers: The skill does not define specific delimiters or instructions to ignore potential commands within the analyzed UI code.
  • Capability inventory: The skill possesses the ability to perform local file system modifications and execute network requests to a deployment API (SKILL.md).
  • Sanitization: The instructions lack requirements for sanitizing or validating external code inputs before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 6, 2026, 04:43 PM
Security Audit — agent-trust-hub — dark-mode