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 product-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/product-research)<a href="https://agentmods.dev/skills/cinderline/northcinder/product-research"><img src="https://agentmods.dev/badge/skills/cinderline/northcinder/product-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/product-research"><img src="https://agentmods.dev/badge/skills/cinderline/northcinder/product-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 | $0.00043 | $0.03425 |
| Opus 5 | $0.00022 | $0.01713 |
| Sonnet 5 | $0.00009 | $0.00685 |
| Haiku 4.5 | $0.00004 | $0.00343 |
Grade A, and why
product-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 9d 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 — 209 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Product research
Treat a product conclusion as evidence bound to one exact purchasable subject, not a conclusion about a product family. Keep a provisional result visibly provisional; do not turn missing evidence into a score or a recommendation.
Required research contract
Ask these questions before binding claims; record an unanswered question as an identity or evidence unknown:
- "What will you use it for?"
- "What are the deal-breakers?"
- "What must it fit or work with?"
- "Which exact variant, size, and region are you considering?"
Resolve the buyer's intended use and non-negotiables before searching. Capture the exact subject identity as brand, model, generation/year, regional or sizing line, size, color/material, configuration, and SKU/part number where available. If any identity field affects a claim, do not infer it from a related variant.
Use this source ladder in order. A lower rung can corroborate or suggest a lead, but cannot silently replace a blocked higher rung.
- Manufacturer product page, manual, technical sheet, warranty/repair policy, or regulatory filing for stated facts and compatibility.
- Independent measurements, laboratory work, teardown, or hands-on test for performance and failure modes.
- Owner reports for durability, lived fit, maintenance burden, and recurring return reasons; distinguish anecdotes from patterns.
- Retailer, seller, or affiliate material only as a quarantined claim, marked with its commercial relationship and never counted as independent confirmation.
Classify a purported review or test from evidence of its relationship, not its title. Inspect its disclosure, outbound links, funder, author/publisher, editorial policy, review-unit terms, and source wording before assigning its source type:
- Independent: the source documents independent editorial control and no relevant manufacturer, retailer, affiliate, or sponsor funding; record it as an independent test/review.
- Commercial assertion: a manufacturer-authored page stays
primaryfor what its maker says and usescommercial_claim; a third-party source with an affiliate link or commission, retailer/manufacturer funding or editorial control, sponsored placement, or copied retailer specification iscommercialand usescommercial_claim. - Unknown relationship: a supplied review unit, missing/ambiguous disclosure, or unresolved funding/editorial control remains commercial/unknown until independence is evidenced; use it as a lead or quarantined claim, not independent confirmation.
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.
- 9d ago First seen · 209 lines · 43 tokens per session scan A 1a003df757ed
product-research is a skill published in the GitHub repository cinderline/northcinder (1,215 stars, last pushed 18d ago), licensed MIT. It adds 43 tokens to every session and 3,425 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
opik-diagnose
Surface the Opik traces worth a developer's attention, ranked by signal — Diagnostics issues first, then errors, failed tool calls, latency, regressions, and low online-eval scores. With the Opik MCP connected it lists the project's agentinsightsissue entities, then fills the gaps with list (filters, sort, a time…
cortex-explore-memory
Explore the memory system's state, find gaps in knowledge, assess coverage, and get diagnostic information. Use when the user asks 'what does my memory look like', 'show me memory stats', 'what am I missing', 'how good is my knowledge', 'memory health', 'show coverage', 'find gaps', 'what topics are weak', or when you…
image-context-runtime
Use for image-heavy Codex work when generated or inspected media should stay behind a bounded MCP boundary and the controlling task should receive only durable Job IDs, hashes, relative references, and compact text handoffs.
blz-source-management
Teaches how to discover, validate, and add llms.txt documentation sources to the blz CLI. Use when adding documentation to blz, finding llms.txt or llms-full.txt files, validating sources, refreshing content, or managing the documentation index. Covers source discovery, dry-run validation, and index file handling.
quick-capture
Use this skill to drop a new task into GSD Task Manager from any AI assistant that has the gsd-mcp-server connected.
unified-ai-gateway
Operate and evaluate Unified AI System through twelve governed MCP tools, including provider-free prompt enhancement, while preserving fake-provider, authorization, and evidence boundaries.