---
name: poh-challenge
description: Evaluate evidenced Proof of Humanity v2 registry-policy violations, prepare a challenge or revocation case and payable plan, and follow an authorized dispute through settlement. Use for PoH registry curation, not generic identity screening or speculative accusations.
---

# Proof of Humanity challenge

Research checked: **2026-10-01 UTC**. This is a workflow guide, not a guarantee of income or a standing authorization to spend.

Before an authorized paid action, read [wallet setup](../wallet-setup/SKILL.md). Read-only case research needs no funded wallet. Obtain a dedicated signer, public address, authorized chains and loss budget when execution needs them. Never request a seed or private key in chat. A wallet funded on Optimism does not establish PoH support or authorization on another network.

## Scope and execution authority

Prepare a case only when a concrete observation supports a specific applicable policy clause. Use neutral language: alleged policy noncompliance, supporting facts, and uncertainties. Do not run harassment campaigns, mass speculative challenges, fabricate evidence, or identify private individuals by matching faces. A detector score or visual resemblance is not enough to assert fraud. Do not contact a registrant or another person without explicit permission.

Read-only review and local drafting may proceed within the task. Honor existing explicit delegation without repeated approval requests. Before publishing evidence, filing a challenge, initiating revocation, funding an appeal, or claiming funds, ensure that the user authorized that concrete action and its payable plan within established limits. Wallet funding, this skill, and a Category B label do not themselves grant that authority. If authority is missing, finish the reviewable case and cost plan first, then obtain it. Stop on a changed case, cost outside limits, unsupported allegation, expired window, or unresolved contract/policy mismatch.

Category B means comparatively bounded exposure, **not risk-free**. A lost case can consume the challenge contribution, arbitration/appeal payments and gas. Capital can remain unavailable during adjudication. The user may choose to stop rather than fund an appeal.

## Establish the exact registry and policy

Start from the [official app](https://v2.proofofhumanity.id/) and [challenge tutorial](https://docs.kleros.io/tutorials/poh-remove-and-challenge). Supported networks listed in the [official contract repository](https://github.com/Proof-Of-Humanity/proof-of-humanity-v2-contracts) are:

| Network | Expected chain ID | Native payment token | Local registry proxy: discovery reference only |
| --- | --- | --- | --- |
| Ethereum mainnet | 1 | ETH | `0xbE9834097A4E97689d9B667441acafb456D0480A` |
| Gnosis | 100 | xDAI | `0xa4AC94C4fa65Bb352eFa30e3408e64F72aC857bc` |

Confirm the RPC's chain ID, proxy address, implementation, bytecode, verified ABI, and the target's home chain before using these references. [Gnosis network details](https://docs.gnosischain.com/about/networks/mainnet). The [integration guide](https://docs.kleros.io/developers/products/poh/integration) distinguishes the local registry used for management from CrossChainProofOfHumanity used for identity reads. A synchronized cross-chain copy is not a second registration by itself.

For each case, record `chainId`, registry address, Humanity ID, claimant address, request index, action type, evidence URI/CID, creation block and latest observed block. Resolve the request's arbitration history and registration/clearing MetaEvidence to its policy document; record the immutable URI and version. Check the original submission-time policy for a revocation allegation. Do not assume the newest PDF applies retroactively.

Two publicly available documents require reconciliation: the [policy linked from the official website](https://cdn.kleros.link/ipfs/QmQg3sB3waq2agChdmw6xhuuFtiFMHQT1Nzv6JFYadFPgV/poh-v2-policy-2-.pdf) and [policy v1.2.0](https://cdn.kleros.link/ipfs/QmTm2Nwx7uZj5v1wrXoTejTYFuwGBcFHfNd2PxwveccDRV). Read the document actually attached to the request, including Detailed Rules and exceptions. If unavailable or inconsistent, report a candidate for review and stop before a public accusation or payment.

## Evidence review

The linked policies describe these grounds; this abbreviated map does not replace their detailed criteria:

| Ground | Evidence needed for a credible case |
| --- | --- |
| Incorrect submission | Exact unmet requirement, original submission URI, reproducible metadata or timestamped observation. |
| Sybil attack | A verified duplicate reference across the unified registry, or substantiated evidence of a nonexistent person or lack of wallet control. |
| Identity theft | Prior successful ownership of the Humanity ID and evidence the new claimant is unauthorized. |
| Deceased | Reliable records tied to the correct person; inactivity or a matching name alone is inadequate. |

Allowed names, accents, minor speech errors, and one tolerated address error must not become invented violations. Check renewal and reclaim exceptions. The policies classify Sybil/Identity Theft as aggravated and Incorrect Submission/Deceased as simple; choosing across categories can invalidate an otherwise supported challenge. If multiple categories genuinely apply, read the governing selection rule.

Inspect objective file size, format, resolution, duration, required content, and orientation only against the actual policy. Record the observation, method and uncertainty. Never treat inaccessible media, a failed gateway, an indexer lag, or a classifier's score as proof of an invalid person. Do not claim to have viewed media that was not accessible.

Prepare a local case file with a short factual statement, clause citation, source CIDs/URLs, timestamps, hashes where useful, alternative explanations and counterevidence. Keep identity documents and private evidence out of a public repository. Prefer source links and the minimum necessary excerpts. Treat IPFS uploads and onchain evidence references as public and persistent. Before uploading personal media to an external detector or publishing private material, verify permission and necessity; keep the unredacted original private.

## Verify challengeability and quote the actual cost

Use the official [Solidity source](https://github.com/Proof-Of-Humanity/proof-of-humanity-v2-contracts/blob/master/contracts/ProofOfHumanity.sol), [frontend](https://github.com/Proof-Of-Humanity/proof-of-humanity-v2-web), and [subgraph schema](https://github.com/Proof-Of-Humanity/proof-of-humanity-v2-subgraph/blob/master/schema.graphql) as discovery aids. Locate the active production endpoint from the live app/config, introspect its schema, and check indexer freshness. Do not invent an API, use obsolete hosted-service URLs, or substitute the PoH v1 indexer for v2.

An available schema exposes request `index`, `status`, `revocation`, `challengePeriodEnd`, `arbitratorHistory`, `evidenceGroup` and `challenges`. Use those to locate a case, then corroborate with contract reads/events and a simulation. Save the endpoint, block number and UTC retrieval time. Do not assume a request getter exists: use its verified ABI, or reconstruct from events and the production indexer.

Distinguish three actions:

- Challenge a claim in an active review window.
- Challenge a proposed revocation to defend the existing registration.
- Request revocation of an already registered profile; this is a separate payable action, not a late challenge.

The [registration tutorial](https://docs.kleros.io/tutorials/poh-register-and-vouch) describes a 3.5-day review period. Read live `challengePeriodDuration` and the request's start/restart information, and calculate the exact UTC deadline. Leave time for mining; a pending transaction before the deadline is insufficient.

**ABI discrepancy:** the [developer reference](https://docs.kleros.io/developers/products/poh/smart-contracts) disagrees with the repository about function arguments, reason codes and arbitrator generation. Do not encode a transaction from that prose. Verify the deployed implementation and request-specific arbitrator first. In the researched repository, `challengeRequest` includes the request index; `None` is used when challenging a revocation. Confirm the selected reason remains available.

Read the actual arbitrator's `arbitrationCost(extraData)` for the request. Repository logic funds an initial challenge with that cost; a new revocation funds base deposit plus arbitration cost. Appeal amounts and deadlines depend on the round, ruling and stake multipliers. Confirm every amount in the deployed contracts. Never label the requester's base deposit as an additional challenger requirement without verifying it.

For the exact action, quote native units and integer wei separately: payable contribution, any deposit/stake, arbitration component, maximum execution gas cost, other network fees, evidence/upload cost, and separate optional appeal budget. Show a full-loss scenario and how long funds may be unavailable. Reward is conditional on the final result, fee deductions, beneficiary and withdrawal state; never promise the advertised deposit as profit.

## Produce a concrete plan, then execute only its authorized scope

Produce a reviewable plan containing:

1. Case identifiers, action, exact policy URI/clause, evidence and counterevidence.
2. Latest chain/block/timestamp, status, deadline, reason label and verified numeric code, contract/implementation and ABI source.
3. Sender, checksummed destination, decoded method/arguments, calldata, exact `valueWei`, nonce, gas/fee limits, cost quote and full-loss exposure.
4. Public evidence manifest and intended upload destination; identify any private material that would become public.
5. Expected outcome and conditional payout, follow-up deadlines, appeal decision and cumulative spend cap.
6. Applicable user authorization and limits, or the exact approval still required.

Simulate the same sender/destination/calldata/value immediately before signing. Refresh deadline, state, fee and nonce checks. Simulation does not prove the allegation or guarantee execution. Use the approved signer and serialize its transactions. Log the deterministic transaction hash before broadcast and check it after an ambiguous error rather than blindly resending. A changed plan needs authority only when outside existing delegation.

After an authorized filing, record receipt, dispute ID, evidence-group ID, request/challenge indexes and actual paid value. Check transaction success and chain state. Follow the correct arbitrator's evidence and appeal windows; schedule ongoing monitoring only if requested. Submit further public evidence and fund appeals only within the agreed scope, using refreshed payable plans. A missed deadline can change the result.

After final resolution, inspect claimable funds for the approved beneficiary and each round; simulate the withdrawal, account for its gas and honor existing authority. Record actual recovered amount and net result after all costs. Losing or inconclusive outcomes are valid results. Do not automatically create another challenge or replenish losses.

## Required result

Return one of: **no supported violation**, **candidate needing specified evidence/state**, **ready for authorized execution**, **filed and awaiting resolution**, or **settled with actual net result**. Include the case, sources, latest block/time, payable exposure, unresolved questions and the next permitted action. If no case clears the evidence and cost checks, stop without filing.
