Opportunities, Bounties, and Assignments

Draft 0.1 — 2 August 2026
This protocol does not determine whether a participant is an employee, independent contractor, volunteer, prize contestant, vendor, partner, or other legal category. The actual relationship and applicable law control.

Objective

ANUKA should let a person enter through a real opportunity and begin contributing without navigating a conventional hiring funnel.

That convenience must not come from hiding terms, shifting risk to contributors, or inviting unlimited unpaid work.

This protocol defines how a need becomes a legitimate opportunity and how that opportunity becomes an assignment, contribution, decision, payment, and portable record.

Core records

Opportunity record

An opportunity is a published need or invitation.

Required fields:

opportunity_id;
issuing product or company;
responsible owner;
title and plain-language summary;
opportunity type;
desired outcome;
scope and exclusions;
eligibility and assurance requirements;
required tools, access, or data;
expected time range;
application or claim method;
assignment model;
acceptance criteria;
review authority;
reward model and currency;
IP and confidentiality terms;
cancellation and refund rules;
dispute procedure;
legal-relationship notice;
publication and attribution settings;
creation, expiry, and revision timestamps.

Application record

An application is a participant’s request to undertake an opportunity.

It SHOULD contain only data necessary for selection.

Possible fields:

relevant credentials or contribution evidence;
proposed approach;
availability window;
requested budget or terms;
declared conflicts;
data-access requirements;
preferred public attribution;
confirmation of material terms.

Assignment record

An assignment is a confirmed agreement that a specific participant, team, or agent may perform the work.

It MUST freeze or version:

scope;
acceptance criteria;
reward terms;
deadlines;
permitted access;
review authority;
IP terms;
cancellation conditions;
dispute path.

Substantial work SHOULD NOT begin before an assignment exists unless the opportunity is explicitly structured as an open contest or discovery bounty.

Contribution record

A contribution links the assignment to submitted evidence and deliverables.

Decision record

A decision states whether the contribution was:

accepted;
accepted with conditions;
returned for declared revisions;
rejected with reasons;
canceled due to issuer failure;
superseded;
escalated to dispute.

Settlement record

A settlement records the final financial and reputation outcome.

Opportunity types

Fixed-deliverable bounty

A declared reward is paid when acceptance criteria are met.

Suitable for:

integration implementation;
documentation;
design asset;
migration;
reproducible bug fix;
defined test plan;
research brief.

Milestone assignment

The work is divided into independently reviewable and payable stages.

Suitable for larger work where all-or-nothing acceptance would unfairly shift risk.

Discovery bounty

A reward is paid for identifying and documenting a valuable issue, lead, vulnerability, or opportunity.

The issuer MUST define duplicate handling and prior-knowledge rules.

Test or qualification bounty

Participants are paid to test, reproduce, evaluate, moderate, or qualify a release or claim.

Outcome bounty

Some compensation depends on a measured result after release.

Outcome bounties require:

baseline;
measurement window;
source systems;
primary and guardrail metrics;
attribution method;
external-event policy;
minimum evidence threshold;
maximum reward;
downside and neutral-result rules.

Open contest

Multiple participants may submit competing solutions.

A contest MUST state:

number of winners;
judging criteria;
total available rewards;
whether non-winning entries receive compensation;
IP treatment for non-winning work;
deadline;
conflict policy;
tie and cancellation rules.

A contest MUST NOT be disguised procurement of many complete unpaid alternatives.

Stewardship opportunity

A product or domain seeks a time-bounded operator. This is governed by the Product Stewardship Protocol and may require a separate legal agreement.

Assignment models

Claimable

The first eligible participant can claim a low-risk task.

Curated

The issuer selects from applications.

Matched

ANUKA recommends candidates based on declared evidence and constraints. The issuer or participant still decides.

Panel-selected

A qualified panel selects a participant for higher-risk work.

Open parallel

Several explicitly authorized participants work in parallel under contest or paid-exploration terms.

Agent-executable

An authorized AI agent may perform work under a principal, tool, budget, and review policy.

Bounty integrity

A bounty must be real before it is promoted.

The issuer SHOULD demonstrate one or more of:

funded payment intent or approved budget;
connected payment account;
binding internal budget authorization;
sponsor commitments;
declared non-monetary reward with legitimate value.

The interface MUST distinguish:

funded;
budget-approved but not reserved;
sponsor-dependent;
unfunded community request;
non-monetary.

Duplicate work

Duplicate handling is essential for discovery and bug bounties.

The opportunity SHOULD define:

what counts as the same issue;
whether first valid report wins;
whether materially independent reproduction earns partial reward;
how timestamps are established;
whether private disclosure is required;
appeal procedure.

A contributor should be able to check for known duplicates without revealing sensitive private reports.

Acceptance criteria

Criteria must be:

observable;
versioned;
proportionate;
available before assignment;
tied to evidence;
separated from unstated preference.

Criteria may include:

automated tests;
reviewer checklist;
design specification;
deployment status;
measured performance threshold;
security review;
user test;
documentation completeness;
rollback readiness.

The issuer MUST NOT retroactively add material requirements without participant agreement and appropriate compensation adjustment.

Revision cycles

The assignment SHOULD define included revision rounds.

A revision request must identify:

unmet criterion;
evidence;
requested change;
deadline;
whether additional work is in scope;
whether compensation changes.

Unlimited revision language is prohibited.

Cancellation

Contributor cancellation

The contributor may withdraw according to declared terms. The record should distinguish voluntary withdrawal, inability to continue, safety concern, and issuer breach.

Issuer cancellation before assignment

No payment is normally due unless the opportunity terms state otherwise.

Issuer cancellation after assignment

The participant SHOULD receive payment for accepted milestones and may receive a kill fee for committed capacity or completed work.

Platform cancellation

ANUKA may suspend an opportunity for fraud, sanctions, security, legal, safety, or policy reasons. The platform must preserve evidence and explain material consequences where lawful.

Unpaid work safeguards

ANUKA SHOULD NOT require extensive unpaid proposals, custom implementation, or speculative deliverables as a condition of ordinary selection.

Permitted lightweight selection materials may include:

existing work samples;
a short approach statement;
relevant credentials;
limited paid trial assignment;
standardized capability assessment.

If a custom test produces usable business value, it SHOULD be paid.

Access controls

Assignments receive the minimum permissions required.

Every access grant must define:

principal;
resource;
actions;
environment;
start and expiry;
logging;
secrets policy;
review and revocation;
handover or deletion.

Production access is not an ordinary entitlement of winning a bounty.

Contributor teams

An assignment may be accepted by a team.

The team must declare:

members;
accountable lead;
payout allocation;
IP and confidentiality coverage;
agent use;
reviewer conflicts;
what happens when membership changes.

ANUKA should not infer equal ownership or equal payment from team membership.

AI-agent participation

An AI agent may:

discover opportunities;
draft applications;
perform authorized tasks;
generate artifacts;
run tests;
submit evidence;
monitor milestones.

An agent MUST have:

accountable principal;
disclosed identity and version where material;
scoped credentials;
budget limits;
action logs;
human or policy review appropriate to risk;
no right to misrepresent generated work as independently human-authored.

Public opportunity data

Public listings SHOULD expose only information necessary for discovery.

Sensitive details may be released after:

application;
qualification;
NDA or confidentiality acceptance;
owner approval;
step-up identity assurance.

Reputation updates

An opportunity may create portable records for:

assignment;
completion;
acceptance;
release;
outcome;
review quality;
payment status;
dispute and correction.

A rejected proposal MUST NOT automatically become a negative reputation event.

A dispute MUST remain visibly disputed until resolved.

Worker classification boundary

Opportunity design affects legal risk.

Risk indicators can include:

controlling when, where, and how work is performed;
continuous or indefinite relationships;
exclusivity;
setting all prices and methods;
monitoring detailed work behavior;
work integral to the operating business;
economic dependence;
inability to realize profit or loss through independent managerial skill.

Neither a click-through contractor agreement nor a connected Stripe account decides worker status.

ANUKA SHOULD maintain a classification-risk checklist and route high-risk relationships to counsel or an appropriate employment model.

Opportunity state machine

Recommended states:

DRAFT
→ QUALIFYING
→ PUBLISHED
→ APPLICATIONS_OPEN
→ ASSIGNMENT_PENDING
→ ASSIGNED
→ IN_PROGRESS
→ SUBMITTED
→ REVIEWING
→ REVISION_REQUESTED
→ ACCEPTED | REJECTED | CANCELED | DISPUTED
→ SETTLED
→ ARCHIVED

Each transition must record actor, time, reason, and applicable version.

MVP acceptance criteria

typed opportunity and reward models;
funded-status display;
versioned acceptance criteria;
application and assignment records;
milestone support;
minimum-access grants;
submission evidence;
structured review decisions;
kill-fee capability;
duplicate-report policy;
payment and dispute integration;
no hidden unpaid-work funnel;
contributor export.

Sources

Verification record

Sources opened and checked: 2 August 2026
Protocol status: Founding draft
Worker-classification review before launch: Required
State-law review before launch: Required