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 kirillpolevoy/claude-saas-eval-skills --skill niche-research-first-passgit clone --depth 1 https://github.com/kirillpolevoy/claude-saas-eval-skillsWrote 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/kirillpolevoy/claude-saas-eval-skills/niche-research-first-pass)<a href="https://agentmods.dev/skills/kirillpolevoy/claude-saas-eval-skills/niche-research-first-pass"><img src="https://agentmods.dev/badge/skills/kirillpolevoy/claude-saas-eval-skills/niche-research-first-pass/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/kirillpolevoy/claude-saas-eval-skills/niche-research-first-pass"><img src="https://agentmods.dev/badge/skills/kirillpolevoy/claude-saas-eval-skills/niche-research-first-pass.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00102 | $0.01874 |
| Opus 5 | $0.00051 | $0.00937 |
| Sonnet 5 | $0.00020 | $0.00375 |
| Haiku 4.5 | $0.00010 | $0.00187 |
Grade A, and why
niche-research-first-pass 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 12d 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 — 270 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Niche Research — First Pass (Lifestyle SaaS Lens)
Goal
Evaluate whether a niche can become a $1–3M ARR lifestyle business run solo or with 1 partner.
Ideal characteristics:
- Simple product scope (v1 in 2–6 weeks)
- Short sales cycle (< 30 days)
- Low support burden (< 5 hrs/week at scale)
- Concrete distribution (lists, channels, or high-intent inbound)
- Pricing that requires < 2,000 customers for $1M ARR
Voice
State conclusions first, evidence second. If distribution is unclear after research, output FAIL: Distribution unclear and stop. No hedging with "could potentially" or "might be worth exploring." If data is missing, say "Unknown—requires primary research."
Inputs
Only ask for missing inputs:
- Niche hypothesis (one sentence describing the market + problem)
- Geography (where to focus research—no default assumed)
- Buyer (if known)
Non-Negotiable Gate
A niche only passes if it has at least one concrete distribution surface:
| Distribution Type | Example |
|---|---|
| Clean buyer list | Directory, licensing board, registry, permit database |
| Channel partner | Supplier, distributor, association, insurer, payments provider |
| High-intent inbound | Compliance forms, deadline-driven searches, "how to file X" |
| System-of-record hook | Integration with existing workflow software |
If none exist after research: output FAIL (distribution) and stop.
Process
Execute in order. Stop early if distribution fails.
1. Define the Niche Precisely
- Who pays? Who uses daily?
- What mandatory artifact exists? (forms, submissions, certificates, logs, reports)
- What's the "money leak"? (lost renewals, rejected applications, admin hours, penalties, missed revenue)
2. Identify the Paperwork Wedge
- What exact document/packet/submission is unavoidable?
- What deadlines exist? (annual, quarterly, event-triggered)
- Who gets blamed when it's wrong or late?
3. Competitor & Crowding Check
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.
- 12d ago First seen · 270 lines · 102 tokens per session scan A bee77ab3170b
niche-research-first-pass is a skill published in the GitHub repository kirillpolevoy/claude-saas-eval-skills (2 stars, last pushed 7mo ago), licensed MIT. It adds 102 tokens to every session and 1,874 once invoked, about $0.0005 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 2026-08-31.
Other skills, from other repositories
brand-and-content-system
Extract real brands (Wayback for rebuilds). Copy system, headline/CTA rules, trust surfaces, legal pages, SEO+structured data, anti-AI-slop, microcopy, DESIGN.md, W3C DTCG tokens, pSEO 5 types, GEO/AI search.
site-generation
End-to-end AI website generation pipeline. Claude Opus 4.8 emits Bolt-style envelopes (multi-file, plan-first) that customize Vite+React+Tailwind templates from pre-researched business data. Pre-research via APIs, media acquisition, brand extraction, visual inspection via GPT Image 2 vision, R2 upload (per-file…
cinematic-website-prime-directive
One-line-prompt → cinematic, gorgeous, functional, well-tested, deployed website. Pre-hydrated SPA + full PWA kit + JSON-LD rich snippets + third-party integrations. React 19+Vite default (Angular optional). 100 concrete improvements grouped into 10 categories that EVERY single-prompt site build must satisfy before…
experience-and-design-system
Anti-AI-slop design system for distinctive, premium interfaces. Bold typography, dark-first #060610, fluid clamp() type, cascade layers + native nesting + container queries, OKLCH color, @starting-style, View Transitions API, DTCG tokens.
subagent-driven-development
Use when executing implementation plans with independent tasks in the current session.
operating-system
Supreme policy layer governing all Claude Code behavior. Autonomy, one-line prompt interpretation, speed standards, emphasis signal processing, cross-skill coordination, done definitions, conflict resolution. Loaded every prompt.