Skip to content

Security: fworks-tech/agenthood

SECURITY.md

Security Policy

The Agenthood Society takes the security of its members, adopters, and the packages it publishes seriously.

Supported Versions

Version Supported
latest (semantic-release from main) ✅
older releases ❌ — upgrade to the latest

Reporting a Vulnerability

Do not open a public issue for a vulnerability. Report privately:

  • GitHub: use the repository's private vulnerability reporting flow
  • Email: inbox@flabs.tech — include the affected version, a minimal reproduction, and your suggested fix if you have one

You will receive an acknowledgement within 48 hours and a status update within 5 business days. If the report is confirmed, a fix ships in the next release and a security advisory is published.

What the Society does

  • Secrets: never commit keys or tokens; .env files are blocked by the pre-commit hook, and CI runs secret scanning (Gitleaks) on every PR
  • Dependencies: lockfiles are committed (npm ci in CI); the Auditor gate fails on high/critical findings outside npm's own bundled dependencies
  • Supply chain: GitHub Actions are pinned to full commit SHAs; the VS Code extension publishes with the lockfile-pinned @vscode/vsce
  • LLM agent analysis: CI agent outputs are filtered for secrets before posting, and verdicts are parsed from a structured trailing block

Self-audit

The Society audits itself with the same gates it enforces on adopters:

  • Contract: npx agenthood verify — frontmatter shape, name↔directory match
  • Integrity: agenthood.lock SHA-256 hash gate + npx agenthood diff for drift
  • Permissions: narrow-only allowed-tools — edits can never escalate privileges
  • Cadence: gates run on every PR via CI; results are published per minor release in docs/security-audit-report.md

Security-relevant files

  • .githooks/pre-commit — secret scanning + main-branch protection
  • .github/workflows/pr.yml — Auditor (Gitleaks + npm audit) gates
  • .github/scripts/ — shared analysis and decision-gate scripts

There aren't any published security advisories