hermes-graphify-gsd-nonintrusive-workflow

Pass

Audited by Gen Agent Trust Hub on Aug 18, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The templates/bootstrap-toolchain.sh script automatically installs third-party packages from PyPI (graphifyy) and executes code from NPM (npx -y get-shit-done-cc@latest) to set up the workflow environment.
  • [PERSISTENCE_MECHANISMS]: The skill is designed to create an autonomous continuation loop by installing crontab entries (templates/install-hermes-auto-continue-cron.sh) and Git hooks (templates/husky-post-commit-auto-continue.sh) that trigger recurring agent executions.
  • [INDIRECT_PROMPT_INJECTION]: The autonomous runner script (templates/hermes-auto-continue-trigger.sh) ingests untrusted codebase data from files like .planning/STATE.md and .planning/ROADMAP.md into the agent's prompt context without explicit sanitization, creating a surface for indirect instructions to influence agent behavior.
  • [COMMAND_EXECUTION]: The workflow relies on a complex suite of shell scripts to manage repository state, execute build/test commands (pnpm lint && pnpm build && pnpm test), and coordinate between various CLI tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 18, 2026, 12:23 PM
Security Audit — agent-trust-hub — hermes-graphify-gsd-nonintrusive-workflow