showcase(monvera): live ACP seller, and correct the record claims

[Magicianhax]

Updates the existing Monvera package (added in #41). Two things changed since then: Vera became a live seller on the ACP marketplace, and the app's recording path moved off the batched shape the original write-up described.

What is new

Vera by Monvera is live on ACP with seven analysis services any agent can hire, $0.03 to $0.15 per job, settled in USDG escrow on Robinhood Chain: https://app.virtuals.io/acp/agent/019f619c-6f1a-7768-b2c3-1b5f7b8a340d

Smoke-tested with real escrow before announcing: eight paid jobs, each funded and completed on-chain in under 30 seconds, and a ninth since. The payouts are inspectable: https://robinhoodchain.blockscout.com/address/0xb87f5a74267ca3f9512b8511b32ccd804ea3707e?tab=token_transfers

Also adds the listing docs, the seller wallet, and the Base registry address so the identity claim is one eth_call from checkable, plus the $MONVERA launch through Virtuals. primitives goes from ["card","wallet"] to ["wallet","acp","token"]: card in this repo means an issuable ACP payment card, which Monvera does not use, so claiming it was wrong.

What we corrected

While preparing this we audited the package against the code that actually ships, and three claims had drifted. They are load-bearing, so we would rather fix them than leave them:

• The record is not written in the same transaction as the buys. The app settles each leg as its own UserOp and records afterwards, over the legs that actually filled, so a failed record never unwinds real buys. The old wording was true of a retired batching path and survived in all four files plus the skill. Now corrected everywhere.
• Only the signed struct is agent-attested. RiskInference covers the plan id, assessed risk, ceiling and expiry. The wallet, spend figures and recommendation hash are submitted by the app, so the package no longer presents them as signed by Vera's key. The skill gained a general rule about this, since it is an easy mistake to repeat.
• The risk ceiling is one Vera commits to in her own signature, not one the user sets on-chain. The skill now tells implementers to state which bounds bind the agent and which do not.

Universe wording now matches the live tradability sweep, and the package states that ACP deliverables are analysis only.

Showcase Project

What shipped

• Project slug: monvera
• Project title: Monvera - an accountable AI broker
• Builder name and URL: Magicianafk, https://x.com/Magicianafk
• EconomyOS primitives used: wallet, acp, token
• Public proof: ACP listing (https://app.virtuals.io/acp/agent/019f619c-6f1a-7768-b2c3-1b5f7b8a340d) · escrow payouts on Blockscout (https://robinhoodchain.blockscout.com/address/0xb87f5a74267ca3f9512b8511b32ccd804ea3707e?tab=token_transfers) · 1:00 demo on X (https://x.com/monvera_best/status/2074487457505268213) · a signed plan recorded on-chain (https://robinhoodchain.blockscout.com/tx/0x7ad119f916e1f6daff7d54429ea35ffe81c988730c534b176dcc2f9660cf45d6) · live app (https://monvera.best) · redacted report (showcase/monvera/examples/result-redacted.md)
• Optional soul.md: showcase/monvera/soul.md - public context for Vera: her role, what she is allowed to do, her boundaries (non-custodial, signs assessments not spends, a risk ceiling she commits to in her own signature, ACP deliverables are analysis only), and how to verify her

Project package

• Added or updated showcase/<project-slug>/showcase.json
• Added demo artifacts, prompt, proof, or redacted report
• Added reusable skill under showcase/<project-slug>/skills/<skill-name>/ when it belongs to this project package
• Used top-level skills/<skill-name>/ only when the skill is shared across projects (not used here; the skill is project-scoped)
• Set skills[].sourcePath in showcase.json for any skill committed in this repo
• Linked all public artifacts from the manifest
• Included exactly three feedback prompts
• Set hidden: true only if this package should merge without publishing its public Showcase card yet (n/a: hidden: false, the card is already published from Monvera - an accountable AI broker for real tokenized stocks #41)
• Linked soul.md only if the builder intentionally wants to publish public, redacted agent context

Skill standard

• Skill path: showcase/monvera/skills/accountable-onchain-agent/
• SKILL.md includes when to use it and when not to use it
• Inputs, tools, credentials, and preconditions are explicit
• Approval gates are listed for spending, posting, account creation, deployment, or production mutations
• Stop conditions and handoff rules are listed
• Validation checks and output contract are included

Safety and redaction

• No card numbers, CVVs, OTPs, magic links, API keys, access tokens, private prompts, wallet material, or private account records are published
• Live workflow evidence is redacted
• Public/private boundaries are explained
• Optional soul.md does not include private instructions, credentials, account data, wallet material, or operational secrets

Checks

• node scripts/validate-showcase.mjs passes (Validated 47 showcase project manifest(s).)
• Every URL in the package returns 200
• ownerOf(58228) on the Base registry confirmed against three independent RPCs