ghost-scan-secrets

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The initialization agent executes an installation shell script fetched from the author's official GitHub repository to setup the scanner binary.
  • [EXTERNAL_DOWNLOADS]: Downloads the poltergeist binary from the vendor's GitHub releases to perform local analysis.
  • [INDIRECT_PROMPT_INJECTION]: The analysis subagent ingests and evaluates code from the repository being scanned, which could contain adversarial instructions intended to trick the AI into ignoring real secrets or reporting false positives.
  • Ingestion points: agents/analyze/analyzer.md reads source code from the target repository files.
  • Boundary markers: The prompt provided to the analysis agent does not use specific delimiters to isolate the untrusted code content.
  • Capability inventory: The skill utilizes shell execution and automated task dispatch via subagents.
  • Sanitization: The skill redacts detected secret values before including them in finding reports and summaries.
  • [COMMAND_EXECUTION]: Executes the poltergeist scanner binary via the shell to identify potential security risks in the codebase.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 02:28 PM
Security Audit — agent-trust-hub — ghost-scan-secrets