HugoScore hugoscore.org

Full review

Your Healthcare Agent CAIHL draft report

Evidence-linked HugoScore draft report for a health AI tool that affects patients.

Agency checklist · October 30, 2026

Expands 10 of 12 yes

Document-based draft. Independent confirmation is incomplete. Hugo has not personally verified this service.

Assessed function: Open-source CLAUDE.md behavioral spec the patient pastes into their own Claude.ai Project, which turns it into a coverage, billing, appeal and records navigation agent.

Initial Anthropic coding has an Anthropic model in the role record.

A no can mean missing documentation or institution-dependent control; it does not by itself establish harmful behavior. These categories are not clinical safety or product-quality ratings.

  • Public documents do not establish actual product behavior.
  • Stored quotation missing for yes answers: 5. Evidence repair is pending.
See all twelve answers and their evidence
QuestionAnswerEvidence / reviewSource and reasoning
1. Told it is AIYesdocumented
Confirmed at document level
An AI agent for your healthcare. Source

The start page also says the spec 'turns a general-purpose AI into a healthcare agent'. The patient works inside claude.ai, so the AI is evident.

2. Operator and model maker namedYesdocumented
Confirmed at document level
turns Claude (Anthropic's AI) into a knowledgeable, rights-aware healthcare navigation agent Source

The About page says 'Designed and built with love by Nick Dawson in Bozeman, Montana.' (https://www.yourhealthcareagent.org/about/). The patient runs it in claude.ai, hosted by Anthropic. The spec author and the model maker are both named.

3. Patient chooses to use itYesdocumented
Confirmed at document level
The whole agent is a single block of instructions you paste into Claude. Source

The patient sets it up and pastes it themselves. The start page also says 'Many community organizations are also setting this up on behalf of the people they serve.' That is a deployment that could be applied to a patient, but it is not the documented default.

4. Patient sets the taskYesdocumented
Confirmed at document level
Now just describe your situation in plain language. Source

The patient sets the task. Caveat: the spec fixes the domain (coverage, billing, appeals, records) and runs a scripted intake that asks about plan type and ERISA status. It also proactively offers a benefits audit. I found no operator goal beyond navigation help, but the scope is set by the spec.

5. Patient can decline or stopYesSelf-started rule; direct refusal wording not documented
Confirmed at document level
Quotation not recorded; evidence repair pending. Source

Yes follows the adopted Q5 rule for a self-started optional tool, rather than an explicit published refusal right. The original coder's note below predates that normalization and is retained as history: No source says the patient can decline or stop and still get the same service. It is an optional self-started tool and nothing in the sources ties care to it. The start page's 'If that's all you need today, you're done' is not the same statement.

6. Patient sees the AI outputYesdocumented
Confirmed at document level
summarize what the document says in terms the person can understand without losing clinical meaning Source

The patient reads the model's output directly in their Claude chat. Nothing is rewritten by a third party.

7. Patient can correct or challengeNonot-documented
Confirmed at document level
Quotation not recorded; evidence repair pending. Source

The spec tells the agent to flag anything 'potentially incorrect' in documents and to encourage verification with authoritative sources. It documents no route for the patient to correct or contest the agent's own output. The Anthropic privacy policy has a correction-request right for inaccurate personal data in outputs, but the task rules limit host terms to Q9 and Q10.

8. Patient can get a copyYesdocumented
Confirmed at document level
Data exports include conversation data and the user data for your account. Source

Individual-user Claude export provides a qualifying copy; deletion of all local or institutional artifacts is not established.

9. Patient can deleteYesdocumented
Confirmed at document level
which will be removed immediately from your conversation history and automatically deleted from our back-end within 30 days. Source

Host terms apply (Claude.ai). The policy page shows Effective Sep 10, 2026. It also lists a right to request deletion of personal data 'subject to certain exceptions'. YHA itself documents no deletion step, and the sources do not explain how Claude Projects are deleted.

10. Training or sale does not override patient choiceYesdocumented
Confirmed at document level
We may use your Inputs and Outputs to train and improve Anthropic AI models, unless you opt out through your account settings. Source

Host terms decide. Opt-out exists, and the same policy says Anthropic does not 'sell' personal data. The opt-out is not complete. Anthropic still uses data flagged for safety review or explicitly reported to it. The default for consumer accounts is training on unless the patient opts out. The YHA start page tells patients to upload insurance cards, EOBs and denial letters and never mentions this setting. It suggests redacting date of birth, address and member ID only.

11. Patient can run it or take it elsewhereYesdocumented
Confirmed at document level
You may use, adapt, and redistribute this work for any purpose, including commercially, as long as you give appropriate credit. Source

The site says 'The code, the spec, and the guides are open source. There is nothing to install.' (https://www.yourhealthcareagent.org/). The README says to 'use CLAUDE.md directly in any Claude Code project or agentic workflow.' The spec is plain text the patient keeps and can paste elsewhere without anyone's permission.

12. No advertising or data businessNodocumented
Confirmed at document level
sharing your personal data for targeted advertising to promote our products and services Source

No retained. Targeted advertising is the deciding reason; consumer training alone belongs in Q10.

Initial coding: Claude Sonnet helpers, compiled by Claude Opus 5.5 (Anthropic). Selective review and corrections: GPT-6 Astra (OpenAI). Editor: Hugo Campos. Acceptance of corrections is not product verification.

Checklist rules · Independent review and limitations · Later evidence follow-up

Earlier report

Your Healthcare Agent: CAIHL reassessment

September 8, 2026. Published AI-assisted draft.

Your Healthcare Agent’s historical profile describes patient-authored instructions for delegated assistance. Current project verification was unsuccessful. The historical 86 remains provisional, with platform conditions and collaborator disclosures preserved.

Scope and agency posture

Historically a copyable Claude navigation specification. Project instructions and host-platform behavior are separate assessment layers.

Posture: Strongly agency-expanding, with platform and verification caveats.

Axis: 86/100, retained provisionally. No numerical recalibration was performed.

Sources and documented findings

  • The frozen June report documents editable, rights-aware instructions and verification-oriented appeal and record workflows, plus Nick Dawson's disclosed relationship to HugoScore. These remain historical findings.
  • Anthropic's current privacy policy permits consumer input/output training unless users opt out, with specified safety and feedback exceptions. The instruction site's openness does not govern this processing. https://www.anthropic.com/legal/privacy

Patient authority

Inference from the documented workflow: The historical editable specification gives users meaningful direction over administrative assistance. It cannot ensure the host model follows every instruction or obtains approval for every consequential action.

Critical capacity

Editorial assessment: Historical guidance connects documents, rights and verification to concrete challenges and drafts. No fresh project or outcome evidence was recovered.

Informed control

Editorial assessment: The platform's privacy conditions remain material. Preserve the collaborator disclosure and distinguish a publicly inspectable prompt from an independently verified service.

Assessment and limits

The design basis for 86 is specific but not freshly confirmed. An unclear current verification disposition is appropriate without reducing a score because pages failed.

Confidence: Low for current project, moderate for historical design.

Remaining uncertainty: Current setup, specification, action approvals and accessibility remain unverified

Published documentation is evidence of stated conditions, not proof of actual implementation. Unknowns did not receive automatic negative points. Funding, sponsorship, and public code do not determine agency by themselves.

Review provenance

  • Reviewer/model: OpenAI Codex / GPT-6.
  • Method: Focused public-source reassessment using CAIHL: patient authority, critical capacity, and informed control. Existing evidence plus one focused primary-source pass and at most one targeted follow-up. No live product testing. Numerical scores remain provisional editorial placements, not a new calculation.
  • Human review: Hugo Campos authorized publication of these AI-assisted draft reassessments on September 8, 2026. This does not claim comprehensive human verification of every finding.
  • Earlier review: 2026-06-22. Historical assessment. Earlier claims are not automatically reverified by this publication.

Disclosures

Related to an already disclosed HugoScore collaborator. The project was designed and built by Nick Dawson, whose OwnChart project is already profiled by HugoScore with a maintainer conflict-of-interest disclosure. Your Healthcare Agent also explicitly cites CAIHL and HugoScore. This profile should be read as a draft and would benefit from third-party review.