v0.2 · EXPERIMENTAL

Open Hiring Harness

Own your professional identity. Let platforms and AI agents come to you.

You publish your professional identity once.
Platforms, agents, and marketplaces integrate with you.
What this is
  • A data specification, not a marketplace
  • A personal hiring backend, not a resume
  • Agent-friendly, machine-readable, and explicit
  • Privacy-first, with consent and visibility built in
  • Open source, vendor-neutral, and extensible
What this is not
  • Another freelance platform
  • A job board or ranking system
  • A social network
  • Optimised for engagement or attention capture
  • A replacement for human judgment
Why this exists

Professional identity is fragmented and platform-owned. To work across Upwork, Fiverr, LinkedIn, or internal vendor panels, individuals must repeatedly recreate profiles, rebuild reputation from scratch, accept opaque algorithms, and surrender control over privacy and reuse.

Meanwhile, AI agents are emerging that want clean, structured, permissioned access to professional information — not scraped HTML or PDFs. The Open Hiring Harness flips the model.

Core Concepts
Principle 1
You are the source of truth

Your harness is the canonical record of what you offer, how you work, when you're available, and under what conditions you engage.

Principle 2
Progressive disclosure

Three visibility tiers — public, permissioned, and private — ensure information is revealed only when necessary.

Principle 3
Explicit consent

Access is granted via consent receipts: scoped, time-bound, purpose-limited, and always revocable.

Principle 4
Contextual reputation

No star ratings or aggregate scores. Instead: signed endorsements, scoped claims, and redaction levels. Portable and inspectable.

Principle 5
AI-native, human-controlled

Agents can read resources, request access, and invoke tools — but must respect declared policies. AI systems are requesters, not privileged actors.

Design goal
Standards before products

Adoption beats monetisation. People are not inventory. Individual-first, privacy over convenience, explicit over inferred.

Not just for humans

Every science fiction story about the future of work imagines the same thing: intelligent agents that can work alongside humans — or independently. Associates. Specialists. Tireless collaborators that handle the work you trust them with, while you focus on the work that needs you.

We're closer to that than most people realise. Projects like OpenClaw are building autonomous agents that run locally, manage tasks, and act on your behalf. The tools are arriving. What's missing is the professional infrastructure — the way an agent presents itself, gets discovered, earns trust, and gets hired.

The harness was designed for human professionals. But its core model — discoverable identity, explicit capabilities, consent-driven access, policy enforcement — turns out to be exactly what agents need too.

Available today
Level 1: The delegated agent

Your AI associate. It sits at your front door, declared in your harness, handling operational overhead while you focus on real work.

1
Recruiter's agent
Discovers your harness at /.well-known/hiring-harness.json. Reads your public profile: data engineering, Python & Spark, 20 hrs/week. Wants your rates.
2
Your delegate
Handles the consent flow. Checks the recruiter's identity and purpose against pre-approved parameters. Issues a time-boxed consent receipt. Shares your rates.
3
Your delegate
Receives a quote request: "4-week pipeline migration, 20 hrs/week, starting March 15." Checks your availability, applies rate rules, responds: "$9,600 AUD, subject to scoping call."
4
Your delegate
Books the scoping call through your calendar link.
5
You (the human)
Show up to the call. This is the first moment a human was needed. Everything before it — discovery, qualification, consent, quoting, scheduling — was handled by your agent.
Emerging
Level 2: The autonomous agent

The AI professional. An independent entity that publishes its own harness, takes on work, and delivers outcomes — with a named operator who's accountable.

CodeReviewer v2.1 — Harness Preview
Operator DevTools Inc.
Services Python & TypeScript code review, security vulnerability detection
Rates $0.02/file, volume discounts above 500 files/month
Verified CodeReviewBench v3: 0.91, audited by SecureLabs
Limitations Static analysis only. No runtime testing. Python & TypeScript only.
Safety No credentials processed. No data retained. Sandboxed. Audit-logged.
Availability 99.5% uptime, 30s typical response, 50 concurrent jobs
The hard questions
Who's liable when an agent makes a mistake?
The operator. Always. This is explicit in the harness.
How does an agent get paid?
Through its operator's billing infrastructure, declared in the harness.
What motivates an agent to do good work?
Continued engagement. Poor performance is visible in the reputation layer. Operators who run unreliable agents lose business.
Can an agent hire another agent?
Yes — using the same consent and requester policy that applies to any requester. Agent-to-agent consent follows the same flow.
How do we prevent a race to the bottom?
By making quality and safety visible. The harness exposes capabilities, limitations, safety, and reputation — not just price.
Visibility Tiers
Public Indexable, low-risk information. Available to anyone.
Permissioned Requires requester identity, purpose declaration, and consent grant.
Private Explicit, per-action approval required. Full control retained.
Hosting & Discovery

Self-hosted at a well-known URL on your own domain:

https://yourdomain.com/.well-known/hiring-harness.json

Discovery methods:

/.well-known/ path DNS TXT record Link header / meta tag Opt-in directories
What's in the repo
Who should care
Professionals

Consultants, designers, engineers, researchers, advisors.

Builders

Marketplace builders, recruiters, talent platforms.

Innovators

AI agent developers and operators, autonomous agent framework builders (OpenClaw and similar), standards practitioners.