agentleFS
SafetyMPSafetyMP/HR-ERP/AGENTS.md

HR-ERP

AGENTS.md1 starsChanged 6 months agoApache-2.0View on GitHub

This repository is an HR scaffold for testing agent governance (T0–T4) on a multi-tenant SaaS fixture. ESS, payroll, and benefits are a fixture domain so hooks have realistic blast radius. This is not a certified payroll vendor and not a production HRIS. Corporate/site factory overlay (gates, non-self-approval, site id hr-erp): docs/factory-overlay.md. September 2026 pivot: docs/DESIGN-PIVOT.md. Full e2e parity needs Postgres and demo seed — docs/QA.md. CI uses Node 22 + npm 10 (packageManager: npm@10.9.2); do not regenerate package-lock.json with npm…

The file

We split this AGENTS.md into the rules it gives an agent

6 rules. About 0 are common, 4 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Never treat this tree as production payroll, legal HR advice, or a company HRIS.

    line 23
  2. Distinctive:

    Never put real employee, tax, or payroll PII in fixtures, issues, or repros — synthetic data only.

    line 24
  3. Very specific:

    Never claim a green gate from prose.

    line 25
  4. Distinctive:

    Never duplicate Cedar, receipt ledgers, or sandbox runtimes here.

    line 26
  5. Very specific:

    Never set NEXT_PUBLIC_ALLOW_DEMO_DEV_SIGNIN on Vercel Production.

    line 27
  6. Distinctive:

    Before new capabilities, read:

    line 45

≈ “Common” is an estimate: at least 1 in 5 of the 41,581 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
AGENTS.md

HR ERP — community agent contract

This repository is an HR scaffold for testing agent governance (T0–T4) on a multi-tenant SaaS fixture. ESS, payroll, and benefits are a fixture domain so hooks have realistic blast radius. This is not a certified payroll vendor and not a production HRIS.

Corporate/site factory overlay (gates, non-self-approval, site id hr-erp): docs/factory-overlay.md.
September 2026 pivot: docs/DESIGN-PIVOT.md.

Commands

CommandPurpose
npm run governance:lintTier + path classification
npm run governance:planDelegation / handoff plan
npm run governance:ciMerge-bar governance gate
./scripts/harness/verify.shDefinition of Done (lint, governance, build, unit tests)
./scripts/verify.shOptional human wrapper (same gate; outside digest boundary)
./scripts/adversarial.shAuthorized local adversarial probes only

Full e2e parity needs Postgres and demo seed — docs/QA.md. CI uses Node 22 + npm 10 (packageManager: npm@10.9.2); do not regenerate package-lock.json with npm 11.

Never

  • Never treat this tree as production payroll, legal HR advice, or a company HRIS.
  • Never put real employee, tax, or payroll PII in fixtures, issues, or repros — synthetic data only.
  • Never claim a green gate from prose. Run ./scripts/harness/verify.sh (or the documented wrapper) and keep the output.
  • Never duplicate Cedar, receipt ledgers, or sandbox runtimes here. Pair FidusGate for runtime receipts.
  • Never set NEXT_PUBLIC_ALLOW_DEMO_DEV_SIGNIN on Vercel Production. Production JWT mint requires ALLOW_PRODUCTION_JWT_MINT=1 (human authorization). Demo preview on a Production-only deploy also needs ALLOW_DEMO_PREVIEW_ON_PRODUCTION=1.

Cross-tenant isolation is proved by tests/integration/security/employee-rls-cross-tenant.integration.test.ts in the CI integration job. The test fails if a session bound to tenant B can read tenant A's employee by primary key. ./scripts/verify.sh clears DATABASE_URL and does not run that suite.

Agent guide (project)

Repository role: Teaching kit / governance harness on an HR fixture. Not a certified payroll vendor — docs/DESIGN-PIVOT.md, docs/meta/evergreen-open-source-positioning.md.

Team roster: docs/meta/agent-team-map.md.

Project skills: .cursor/skills/README.md · community copy of data-custody: .github/skills/hr-data-custody/SKILL.md · site delivery: .cursor/skills/site-delivery/SKILL.md.

Global skills (lazy): ~/.cursor/skills/README.md · @skill-router at T1+ per ~/.cursor/rules/core-dynamic-skills.mdc.

Team MCP (IDE plane): .cursor/mcp.json — context7, prisma. Product copilot MCP is separate (lib/copilot/).

Always-on boundaries: .cursor/rules/repo-boundaries.mdc + docs/architecture/.

Before new capabilities, read:

  • Design pivot — compete on T0–T4 governance, not HR features
  • Stakeholder value plan — fixture walk paths (W1–W7), not a buyer HRIS roadmap
  • HR Product Owner operating model
  • Feature briefs: docs/product/feature-briefs/ · template
  • Compliance: docs/compliance/ — load @hr-regulated-domain (T3)
  • Payroll kernel (fixture): packages/payroll-calc/ — L3 in hr-regulated-domain/references/
  • Migrations: database-migrations-and-state.md — @hr-data-custody (T2)
  • AI governance: docs/ai-governance/ — @hr-regulated-domain (T3)

Orchestration

Native runtime: cursor-3-native-runtime.md · prefer Auto-review Run Mode · /multitask, /worktree, /best-of-n · hooks + governance:ci / ./scripts/verify.sh as the deterministic merge bar.

Sequence: .cursor/rules/orchestrator-hr-erp.mdc · ADR 0011.

Harness scripts: npm run governance:lint → npm run governance:plan → npm run governance:ci (hook entrypoint archived under _archives/harness-v4/; shared hook libs remain under .cursor/hooks/ for governance scripts).

Phase 2 evidence: npm run governance:evidence · npm run governance:cloud-session · ADR 0019

Coding standards (September 2026)

Path-specific Copilot instructions: .github/instructions/. Repository-wide Copilot file: .github/copilot-instructions.md. Copilot code review skill: .github/skills/code-review/SKILL.md.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.