startup-audit

Pass

Audited by Gen Agent Trust Hub on Jun 19, 2026

Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [SAFE]: The skill implements a robust secret-redaction policy for its analysis. According to references/signal-extraction.md, the signal extraction process is specifically designed to capture only key names and presence from configuration files like .env, while strictly forbidding the recording of any actual secret values.
  • [PROMPT_INJECTION]: The skill processes untrusted data from codebase files and live URLs, which creates a potential surface for indirect prompt injection. This risk is managed by the skill's architecture, which relies on deterministic mapping of code structures (e.g., dependency manifests and database schemas) to business models, rather than executing or interpreting natural language instructions within the data.
  • [DATA_EXFILTRATION]: Safety is prioritized in the reporting functionality. The references/dossier-html-template.md ensures that the generated HTML dossiers are self-contained with zero network dependencies (inline scripts and styles), preventing the possibility of tracking or data exfiltration via external resource requests when the report is viewed.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 19, 2026, 07:12 AM
Security Audit — agent-trust-hub — startup-audit