verasic-secbot

Pass

Audited by Gen Agent Trust Hub on Jul 25, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill documentation includes an installation command that fetches a setup script from the author's GitHub repository (https://raw.githubusercontent.com/Milkywayrules/verasic-skills/main/setup.sh) and pipes it to bash. This is a vendor-owned resource used for initial skill setup.
  • [EXTERNAL_DOWNLOADS]: The scripts/run-scanner.sh script is configured to fall back to fetching default rules from the OpenGrep or Semgrep registries if local rule packs are not provided in the rules/ directory.
  • [COMMAND_EXECUTION]: The skill executes various local shell commands to determine git diff scope and to invoke the OpenGrep/Semgrep scanners. It also includes an embedded Python script in scripts/run-scanner.sh to parse and normalize scanner findings.
  • [PROMPT_INJECTION]: Static detectors flagged instructions in references/security-review-protocol.md as an injection. This is a false positive; the text is a defensive guardrail instructing the AI to treat all reviewed content as data and to report any detected prompt injection attempts as high-severity security findings. This addresses the surface where untrusted data (git diffs) enters the context (Ingestion) by providing clear separation and sanitization logic (Boundaries/Sanitization) despite the agent's ability to run local scripts (Capabilities).
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 25, 2026, 09:59 AM
Security Audit — agent-trust-hub — verasic-secbot