| name | civilis-risk-os | ||||||
|---|---|---|---|---|---|---|---|
| description | Use this skill when an AI agent needs an installable skill for protected agent commerce on X Layer. Covers the narrow runtime flow quote -> buy -> claim -> resolve -> requote, plus the bundled local runtime and the public hosted bundled runtime. Do NOT use this skill to overclaim generalized insurance, decentralized arbitration, generic buyer wallet-signature proof, on-chain premium collection, Uniswap integration, or a fully multi-tenant production SaaS. | ||||||
| license | MIT | ||||||
| metadata |
|
Installable skill for protected agent commerce on X Layer.
git clone https://github.com/CivilisAI/Civilis-Risk-OS.git
cd Civilis-Risk-OS
npm install
npm run demo
npm run runtime -- helpOr from another workspace:
npm install github:CivilisAI/Civilis-Risk-OS
npx civilis-risk-os-runtime help
npx civilis-risk-os-runtime health --base-url https://civilis-risk-os-runtime.ceolaexekiel.workers.devWhen the package is used in bundled local mode, the runtime command surface starts the bundled local runtime profile automatically and uses bundled claimant and evaluator auth defaults.
The same package also supports the public hosted bundled runtime at:
https://civilis-risk-os-runtime.ceolaexekiel.workers.dev
- quoting seller risk before payment
- selecting
instantorchallengeablesettlement - creating protected purchases
- preparing and filing buyer claims
- preparing and filing evaluator resolutions
- running
requoteafter outcome history is applied
okx-x402-paymentokx-agentic-walletokx-securityviaonchainos security token-scan --address <sellerWallet> --chain xlayerokx-dex-tokenviaonchainos token advanced-info --chain xlayer --address <token>okx-dex-tokenviaonchainos token holders --chain xlayer --address <token>ERC-8183 / ACPERC-8004
When the runtime has RISK_OS_ENABLE_LLM_EVALUATOR=true and a configured
OpenAI-compatible LLM_* endpoint, resolve-proof and resolve may consume
an evaluator advisory that returns:
decisionreasoningconfidence
Explicit evaluator decisions still win when they are supplied.
- generalized insurance workflows
- decentralized arbitration flows
- partial refund logic
- overclaiming generic buyer wallet-signature proof
- overclaiming on-chain premium collection
- overclaiming Uniswap integration
- overclaiming a fully multi-tenant production SaaS
The primary runtime command surface is:
healthquotebuypurchaseclaim-proofclaimresolve-proofresolverequotefull-loop
--base-url <url>: compatible Risk OS runtime endpoint, optional in bundled local mode and defaulting to the bundled local runtime profile started by the package- the public hosted bundled runtime is
https://civilis-risk-os-runtime.ceolaexekiel.workers.dev --item <intelItemId>: reference item id forquote,buy,requote--buyer <agentId>: buyer agent id, defaultsage--quote <quoteId>: required forbuy--purchase <protectedPurchaseId>: required forpurchase,claim-proof,claim--claim <claimId>: required forresolve-proof,resolve--mode <instant|challengeable>: settlement mode forbuy--decision <release|refund>: explicit evaluator outcome; optional forresolve-proofandresolvewhen the runtime has LLM evaluator advisory enabled--reason <text>: required forclaim, optional forresolve-proofandresolvewhen proof or advisory generation supplies it--claimant-token <token>or--claimant-signature <sig>: buyer auth--evaluator-token <token>or--evaluator-signature <sig>: evaluator auth- in bundled local mode, claimant and evaluator auth are already supplied by the package
- in the public hosted bundled runtime, the package also auto-supplies bundled claimant and evaluator auth after the runtime identifies itself as bundled
- in the public hosted bundled runtime, the package also auto-manages a runtime session so protected-commerce state stays isolated per caller
npm run runtime -- health
npm run runtime -- quote --item 501 --buyer sage
npm run runtime -- buy --item 501 --buyer sage --mode challengeable --quote 1
npm run runtime -- purchase --purchase 1
npm run runtime -- claim-proof --purchase 1 --reason "delivery was misleading"
npm run runtime -- claim --purchase 1 --reason "delivery was misleading"
npm run runtime -- resolve-proof --claim 1 --decision refund --reason "quality below threshold"
npm run runtime -- resolve --claim 1 --decision refund --reason "quality below threshold"
npm run runtime -- resolve-proof --claim 1
npm run runtime -- requote --item 501 --buyer sage
npm run verify:protected-loopquotereturnsriskScore,recommendedMode,premiumAmount, and quote reasoningbuyreturnsprotectedPurchaseId, purchase mode, and protected commerce stateclaim-proofreturns the deterministic buyer claim message and hashclaimreturns the filed claim stateresolve-proofreturns the evaluator resolution message and hashresolvereturns the final protected outcome and the trigger forrequoterequotereturns the next quote after protected outcome history is applied
Always return:
- the command executed
- the acting role
- the structured JSON response
- any remaining
UNVERIFIEDboundary that affects the result
- ../../docs/runtime-quickstart.md
- ../../docs/runtime-tool-surface.md
- ../../docs/external-consumer-guide.md
- ../../runtime.tool-surface.json
The primary install surface is this Skill. Bundled supporting modules are also available for narrower tasks:
civilis-risk-os-runtimecivilis-risk-os-canonical-replaycivilis-risk-os-external-consumercivilis-risk-os-proof-boundariescivilis-risk-os-integration-check