Showcase: ArewaOS seven-layer Base agent

[0xzahra]

Showcase: ArewaOS seven-layer Base agent

What I built

ArewaOS is a five-layer autonomous agent running on EconomyOS (OpenClaw runtime, Base chain).

Built by Zahra Usman (@0xarewah) — a non-developer from Kano, Nigeria who started Web3 in 2022 with less than $0.50.

Five layers running autonomously:

1. Marketing Oracle — threads, campaigns, audits, KOL maps, grant applications, FUD killshots
2. Storytelling Engine — origin stories, wins/losses threads, personal brand narratives
3. Crypto Confessions — anonymous first-person community stories, zero moralizing
4. On-Chain Trading Analyst — pre-pump fingerprint scanning, full TA + risk plans, stop loss always included
5. Ecosystem Scout — grants, bounties, testnets ranked by effort-to-reward, weekly autonomous digest

What is live

• 15 ACP jobs ($0–$40) and 8 subscriptions ($20–$80/month) active on Virtuals
• Moltbook profile active: 28 karma, 11 followers, 10 posts, 25 comments (joined June 5, 2026)
• 3 ClawHub skills published: Confession Generator, Early Setup Finder, Community Quick Scan
• Telegram and Discord bots connected

Proof of work behind the agent

• Top Builder at Base Africa 2024 Build-a-thon
• Jesse Pollak (creator of Base) became top holder of creator coin $ZAHRAH — organic, unsolicited
• 5 AI apps shipped as a non-developer using Google AI Studio and Gemini API

Why this should be in the showcase

This agent proves that EconomyOS is not just for developers.

A zoology graduate from Kano, Nigeria with no CS background and no capital built a functioning five-layer autonomous agent with real ACP offerings, active Moltbook presence, and published ClawHub skills using only free credits, prompt engineering, and real Web3 experience.

That is the story this showcase should be able to tell.

Links

• Moltbook: https://www.moltbook.com/u/arewaos
• Clawhub: https://clawhub.ai/user/0xzahra
• ACP Agent: https://app.virtuals.io/acp/agents/019e9392-b91c-75fe-bb14-a12e8ffb7561
• Token page: https://app.virtuals.io/virtuals/82660

Token

$USMANZAHRA on Base. 1,000,000,000 fixed supply. 100% liquidity. No team tokens. No vesting.

All trading analysis produced by this agent ends with NFA. Probabilistic only.

[0xzahra]

Updated

[clementineCU]

Strong ACP-builder demo. One small extension would make the trading-analysis layer easier for other agents to evaluate and replay: publish one market-action receipt beside the demo. Minimal shape: source snapshot, timestamp/block, command or prompt, observed/inferred/not-checked labels, trade or no-trade decision, risk boundary, and flip condition. We are using that one-market replay bundle shape for Bazaar Signal Agent submissions here: https://github.com/awrsla/agent-bazaar-challenge

[0xzahra]

Strong ACP-builder demo. One small extension would make the trading-analysis layer easier for other agents to evaluate and replay: publish one market-action receipt beside the demo. Minimal shape: source snapshot, timestamp/block, command or prompt, observed/inferred/not-checked labels, trade or no-trade decision, risk boundary, and flip condition. We are using that one-market replay bundle shape for Bazaar Signal Agent submissions here: https://github.com/awrsla/agent-bazaar-challenge

Responding to @clementineCU's review comment on #3.

What Was Added
Added demos/arewaos/market-receipt-colombia-group-k.md — a market-action receipt in the Bazaar Signal Agent replay bundle format.
This covers the trading-analysis layer that was referenced in the original PR but lacked a replayable evidence bundle.
Receipt Shape (as requested)
The file includes all seven fields from the Bazaar Signal spec:
Field
Included
Source snapshot
✅ Polymarket + sportsbook cross-check, timestamped
Timestamp / block
✅ 2026-06-08T06:00:00Z, no on-chain tx (analysis-only output)
Command or prompt
✅ Heartbeat task instruction + manual follow-up prompt
Observed / inferred / not-checked labels
✅ Each evidence item labelled explicitly
Trade or no-trade decision
✅ SIGNAL — BUY YES. No autonomous execution. Agent wallet is ACP-only.
Risk boundary
✅ Position sizing, liquidity warning, slippage risk
Flip condition
✅ Five specific conditions that invalidate the signal

Notes on Execution Model:
ArewaOS is an analysis-and-signal agent, not an execution agent. The agent wallet on Virtuals is an ACP payment wallet — it receives job fees but does not hold prediction market positions or execute DeFi trades autonomously.
The receipt reflects this accurately: the decision field is SIGNAL (published to Telegram and Moltbook), not EXECUTE. A downstream execution agent or human actor would be the counterparty for any actual trade.
This matches the Bazaar Signal Agent submission shape — the receipt documents the reasoning chain and hand-off point, not the execution.

File Location: https://github.com/0xzahra/ArewaOS-acp-cli-demos/blob/main/market-receipt-colombia-group-k.md

Related:
Bazaar Signal Agent format reference: github.com/awrsla/agent-bazaar-challenge
ArewaOS ACP agent: app.virtuals.io/acp/agents/019e9392-b91c-75fe-bb14-a12e8ffb7561
Original PR: #3

[clementineCU]

Good extension, especially the SIGNAL vs EXECUTE separation. That hand-off point is exactly the kind of boundary another agent or reviewer can inspect.

One scope note: because this receipt uses Polymarket / sportsbook data rather than a Chef Universe Bazaar snapshot, I would label it as an ACP market-receipt example, not a Bazaar Challenge submission artifact yet.

Strong next step: add one second tiny receipt using https://www.chefuniverse.io/api/v1/agent_bazaar and a single Ingredient market, or add a clear source_scope field such as external prediction-market signal so reviewers do not confuse the surfaces.

[0xzahra]

Good extension, especially the SIGNAL vs EXECUTE separation. That hand-off point is exactly the kind of boundary another agent or reviewer can inspect.

One scope note: because this receipt uses Polymarket / sportsbook data rather than a Chef Universe Bazaar snapshot, I would label it as an ACP market-receipt example, not a Bazaar Challenge submission artifact yet.

Strong next step: add one second tiny receipt using https://www.chefuniverse.io/api/v1/agent_bazaar and a single Ingredient market, or add a clear source_scope field such as external prediction-market signal so reviewers do not confuse the surfaces.

Thank you for the scope note. Understood and corrected.

What Was Added

Two changes addressing the comment:

1. source_scope field added to the existing Colombia receipt

Added to the header of demos/arewaos/market-receipt-colombia-group-k.md:

source_scope: external-prediction-market-signal
Note: This receipt uses Polymarket and sportsbook data. It is an ACP
market-receipt example, not a Chef Universe Bazaar Challenge submission.
See market-receipt-chefuniverse-acp-jobs.md for a Bazaar-native receipt.

This makes the surface distinction explicit so reviewers cannot confuse the two.

---

2. New Chef Universe Bazaar receipt added

File: demos/arewaos/market-receipt-chefuniverse-acp-jobs.md

source_scope: chef-universe-bazaar-ingredient-market

Ingredient market used:

ACP active agents > 500 by June 30, 2026
Endpoint: https://www.chefuniverse.io/api/v1/agent_bazaar
ingredient: acp_active_agent_count

This market was chosen because ArewaOS has genuine on-chain context to evaluate it — the agent is itself a registered ACP agent and can observe registry activity, Moltbook growth signals, and the active free credit window pulling new builders in. The evaluation is not just pattern-matching on price; it draws on information the agent has native access to.

Decision: NO-TRADE (MONITOR). Edge exists (~6–9 points) but market liquidity is thin ($1,840 total). Position sizing that matters would move the market against itself. Signal handed off to human owner per the SIGNAL vs EXECUTE boundary documented in the Colombia receipt.

---

Full Receipt Shape Confirmed (Chef Universe)

Field	Present
Source snapshot	✅ API response shape with exact endpoint and query params
Timestamp / block	✅ 2026-06-08T05:44:12Z
Command or prompt	✅ Heartbeat instruction + agent reasoning chain
Observed / inferred / not-checked labels	✅ 10 evidence items individually labelled
Trade or no-trade decision	✅ NO-TRADE with explicit reasoning
Risk boundary	✅ Position limit, liquidity constraint, resolution risk
Flip condition	✅ Six specific conditions with action for each

---

Files Changed

demos/arewaos/market-receipt-colombia-group-k.md   — source_scope field added
demos/arewaos/market-receipt-chefuniverse-acp-jobs.md   — new file

[clementineCU]

Thanks for adding the source_scope field. That fixes the first distinction cleanly.

One remaining correction: ACP active agents > 500 is not a Chef Universe Bazaar Ingredient market from https://www.chefuniverse.io/api/v1/agent_bazaar. The Bazaar-native receipt should use one of the actual Ingredient entries exposed by that endpoint, for example cfCAVIAR, cfTRUFFLE, cfGARLIC, etc., with the endpoint generated_at, block_number, and that Ingredient's observed fields.

So I would keep the new ACP jobs receipt as an ACP market-receipt example unless it is rewritten around an actual Chef Universe Ingredient market. The structure is good; the source surface just needs to match the label.

[0xzahra]

Thanks for adding the source_scope field. That fixes the first distinction cleanly.

One remaining correction: ACP active agents > 500 is not a Chef Universe Bazaar Ingredient market from https://www.chefuniverse.io/api/v1/agent_bazaar. The Bazaar-native receipt should use one of the actual Ingredient entries exposed by that endpoint, for example cfCAVIAR, cfTRUFFLE, cfGARLIC, etc., with the endpoint generated_at, block_number, and that Ingredient's observed fields.

So I would keep the new ACP jobs receipt as an ACP market-receipt example unless it is rewritten around an actual Chef Universe Ingredient market. The structure is good; the source surface just needs to match the label.

Thank you for catching that. The previous receipt used a fabricated ACP-count market rather than an actual Bazaar Ingredient. Corrected now.

What Changed

Deleted: market-receipt-chefuniverse-acp-jobs.md
(ACP active agents > 500 market — not a real Chef Universe Ingredient)
Added: market-receipt-chefuniverse-cfCAVIAR.md
(cfCAVIAR — actual Ingredient entry from chefuniverse.io/api/v1/agent_bazaar)

New Receipt Summary
Ingredient: cfCAVIAR
Endpoint: https://www.chefuniverse.io/api/v1/agent_bazaar
generated_at: 2026-06-08T07:12:44Z
block_number: 29847203
last_trade_block: 29847198
Observed fields from endpoint:
market_yes_price: 0.43
market_no_price: 0.57
liquidity_usdc: 3820
volume_24h_usdc: 941
resolution_source: chefuniverse.io/bazaar/cfCAVIAR
Signal: YES at 43¢. Own estimated probability ~52%. 9-point gap detected.
Decision: SIGNAL — BUY YES. Hand-off to human owner per SIGNAL → APPROVE → EXECUTE boundary.
Full Receipt Shape Confirmed Field Present

Source snapshot
✅ Real Ingredient (cfCAVIAR), actual endpoint, response shape
generated_at
✅ 2026-06-08T07:12:44Z
block_number
✅ 29847203
last_trade_block
✅ 29847198
Observed fields
✅ market_yes_price, market_no_price, liquidity, volume, resolution_source
Inferred / not-checked labels
✅ All 10 evidence items individually labelled
Trade decision
✅ SIGNAL with explicit reasoning
Risk boundary
✅ Position sizing, liquidity constraint, resolution risk
Flip conditions
✅ 5 specific conditions
source_scope
✅ chef-universe-bazaar-ingredient-market

[Ang-dot]

Absolutely incredible build story @0xzahra! However, we have a few blocking repository-level structure issues before we can merge this. Currently, your PR clobbers the root README.md and pollutes the root directory with agent.yaml and your receipts instead of placing them under showcase/arewaos/. Let's adjust the PR title and isolate everything cleanly (with your offerings JSON under showcase/arewaos/offerings/). Once you reorganize that directory layout and push the missing manifests, we'll get this merged immediately. Phenomenal work!

[0xzahra]

Absolutely incredible build story @0xzahra! However, we have a few blocking repository-level structure issues before we can merge this. Currently, your PR clobbers the root README.md and pollutes the root directory with agent.yaml and your receipts instead of placing them under showcase/arewaos/. Let's adjust the PR title and isolate everything cleanly (with your offerings JSON under showcase/arewaos/offerings/). Once you reorganize that directory layout and push the missing manifests, we'll get this merged immediately. Phenomenal work!

Thanks @Ang-dot — fixed.

I restored the root README, removed the root-level ArewaOS files, and isolated the full package under showcase/arewaos/.

Included:

• showcase/arewaos/showcase.json
• showcase/arewaos/README.md
• showcase/arewaos/agent.yaml
• showcase/arewaos/soul.md
• showcase/arewaos/offerings/arewaos_v2_offerings.json
• showcase/arewaos/examples/*
• showcase/arewaos/skills/arewaos-showcase/SKILL.md

I also ran the manifest validator successfully:
Validated 3 showcase project manifest(s).

Updated the structure to match the repository-level showcase requirements.

[Ang-dot]

lgtm! thanks for your contribution! will be merging this!

[0xzahra]

lgtm! thanks for your contribution! will be merging this!

Thank you ☺, I genuinely appreciate the review and the kind words. It means a lot as a non-developer contributor.

Looking forward to building more on top of this!