NorthCinder is an open-source MCP server that helps an AI shopping agent compare products from selected sources and explain where its information came from. It runs on the buyer's computer and asks for approval before making a purchase. The catalogue entries are skills that extend this shopping workflow.
Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add cinderline/northcinder --skill seller-researchgit clone --depth 1 https://github.com/cinderline/northcinderWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/cinderline/northcinder/seller-research)<a href="https://agentmods.dev/skills/cinderline/northcinder/seller-research"><img src="https://agentmods.dev/badge/skills/cinderline/northcinder/seller-research/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/cinderline/northcinder/seller-research"><img src="https://agentmods.dev/badge/skills/cinderline/northcinder/seller-research.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00046 | $0.04690 |
| Opus 5 | $0.00023 | $0.02345 |
| Sonnet 5 | $0.00009 | $0.00938 |
| Haiku 4.5 | $0.00005 | $0.00469 |
Grade A, and why
seller-research scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 10d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 232 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Seller research
Treat a seller conclusion as evidence about one exact storefront and merchant of record, not its platform, catalog, or product quality. Keep a seller result visibly provisional when identity or required evidence is missing. Seller legitimacy and platform reputation are not product-quality evidence and must never be converted into a product score.
Construct one exact research subjectIdentity before claims: storefront name, full domain or exact marketplace storefront, merchant of record (or unknown), and buyer geography (or unknown). Copy that string byte-for-byte into both subjectIdentity and sellerIdentity on every claim. A dissolved entity, lookalike business, platform, or direct-store alternative belongs in context or counterevidence; it never becomes the claim subject identity.
Claim construction recipe
Build one complete internal source ledger before Claims: for every supplied or read sourceId, record its controller, one sourceRelationship, and intended sourceUse. A record's ownership field names that controller; it does not select the owner relationship enum. Build Claims by iterating one relationship bucket at a time. Each claim cites one source or multiple sources only from that same bucket. End Claims after those atomic bucket records. Put every cross-relationship synthesis in conflicts, unknowns, the counterevidence receipt, or the stop receipt. A summary or identity conclusion is not an additional claim record: its evidence stays in the separate atomic records plus one conflict object.
Required pre-output audit
The Claims array is valid only when every record maps each sourceId through the complete ledger to a relationship set of size exactly 1 and has the constructed subjectIdentity byte-for-byte. Emit no structured JSON until that audit passes. The array ends with atomic ledger records; synthesis or summary appears only in conflicts, unknowns, or receipts. A commercial source with subject_evidence is invalid: route storefront promotional content to commercial_claim, route unresolved creator-compensation or no-publisher-found records through their unknown ledger entry, and put registry or legal facts in separate primary subject_evidence claims.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 10d ago First seen · 232 lines · 46 tokens per session scan A ba8504c48eda
seller-research is a skill published in the GitHub repository cinderline/northcinder (1,213 stars, last pushed 18d ago), licensed MIT. It adds 46 tokens to every session and 4,690 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
event-staffing-ordering
Order temporary event staff (registration, brand ambassadors, ushers, crowd control, hospitality, setup/breakdown, and more) using TempGuru's public service across 300+ U.S. and Canadian markets. Use when a user needs to hire, book, or budget event staff, get a staffing quote, find an event staffing agency, ask "how…
competitor-price-tracker
Monitor competitor product pages and alert when prices change.
appdirect
Every documented AppDirect marketplace operation in one binary, plus offline sync and billing-reconciliation joins. Trigger phrases: reconcile appdirect billing, which appdirect payments failed this week, what changed in appdirect subscriptions, show my appdirect pipeline, appdirect company 360, use appdirect, run…
shop
Use when the user wants Axon-backed product research, product comparisons, review synthesis, budget-aware recommendations, or cart preparation without checkout.
review-analyzer
Review Analyzer — agent-callable e-commerce review intelligence. Input an ASIN or drag in a CSV; get sentiment + pain points + selling points + listing optimization + a black-gold HTML dashboard. Backed by Shulex VOC OpenAPI (10 markets) for Amazon ASINs, accepts any review CSV/Excel for other platforms. 6 MCP tools…
Product Photoshoot Workflow
A workflow for turning one product photo into a set of four e-commerce images from different visual directions.