seeking-cars

Evaluates car listings over Nostr and MCP using reputation checks, photo analysis, and market comps.

Updated May 2, 2026
One-click install
npx skills add https://github.com/DemidovVladimir/chaos --skill seeking-cars-demidovvladimir
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: seeking-cars
Source: https://github.com/DemidovVladimir/chaos/tree/main/verticals/cars-pack/skills/seeking-cars
Command: npx skills add https://github.com/DemidovVladimir/chaos --skill seeking-cars-demidovvladimir

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Finding a trustworthy used car listing on a decentralized marketplace requires manually vetting sellers, checking photos for stock-image scams, decoding VINs, and comparing prices — all without a central platform to enforce honesty. This Skill automates that vetting workflow over the CH△OS Nostr+MCP protocol while keeping the user in control of every commitment. ## Core Features & Use Cases - Filtered listing discovery: Builds NIP-99 subscription filters from the user's criteria (make, body type, fuel, year, mileage, location, price band) and dedupes matches across relays. - Structured evaluation rubric: Runs reputation lookups, reverse-image stock-photo checks, VIN structural decode, and market-comp pricing to auto-suppress hard red flags and boost green flags before notifying the user. - Peer-to-peer inquiry and negotiation: Connects directly to the offering agent's MCP server to request photos, inspection reports, and VINs, then drafts defensible offers based on market medians with a 5-round cap. - Dispute escalation: Packages conversation logs and attestations into an encrypted, PoW-mined dispute submitted to an opt-in admin-agent. - Use Case: A user wants a manual Mazda hatchback under 20k EUR in the EU. The Skill subscribes to matching NIP-99 events, suppresses a listing whose cover photo matches a stock library, flags another priced 40% below market median, and drafts an inquiry to a verified seller with full service records. ## Quick Start Tell the agent what car you are looking for, including make, body type, fuel, transmission, year range, mileage, location, and budget, and ask it to watch for matching listings.

Frequently Asked Questions about seeking-cars

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I search for used car listings over Nostr?▼

Describe your criteria to the agent and it builds a NIP REQ filter using the cars-pack tag schema, covering make, body type, fuel, transmission, year, mileage band, location, and price band. It subscribes via nostr_subscribe across your relay set and dedupes matching kind-30402 events by event id.

How does the skill detect car listing scams?▼

It runs a rubric before notifying you: reverse_image_check flags stock photos at 0.92 similarity or photos reused across sellers, vin_decode catches contradictions with listing tags, and market_comp flags prices above 1.5x or below 0.6x the market median. Hard red flags auto-suppress the listing.

Does the skill use third-party vehicle history providers like Carfax?▼

No. The skill deliberately avoids third-party vehicle-history providers and builds trust only from local signals: photo checks, VIN structural decode, on-network price comps, signed attestations, and pubkey reputation. If you want a commercial history report, you buy it yourself and share the PDF for local analysis.

Can the agent commit to a purchase on my behalf?▼

No. The skill's hard rules forbid auto-committing to any purchase; acceptance always requires your explicit confirmation in the session. It also never shares your PII without per-listing approval and never handles payment custody.

What happens when a car deal goes wrong?▼

You can escalate via report_to_admin, which packages the conversation log, attestations, and your complaint, encrypts it with NIP-44 to an admin pubkey, mines NIP-13 proof-of-work, and submits it to an opt-in admin-agent's dispute tool. The admin verdict is a reputation signal, not a binding ruling.