Skip to content

discolike/discolike-skills / skills/build-icp-from-website

build-icp-from-website

Verified

Use when the user asks to "set up my ICP", "build an ICP from our website", or draft an ICP from a best customer's domain. Follow the discolike skill's access and spend rules.

FileKindSize
SKILL.mdskill md1.7 KB
---
name: build-icp-from-website
description: Use when the user asks to "set up my ICP", "build an ICP from our website", or draft an ICP from a best customer's domain. Follow the discolike skill's access and spend rules.
allowed-tools: Read, Grep
---

# Build an ICP from a website

Follow the `discolike` skill's How to work rules first.

New account, no ICP written down, or "set up my ICP".

1. **Read the source.** MCP `extract-website-text` on the user's own domain, or on their best customer's. CLI `discolike extract https://example.com --format json`. SDK `client.companies.extract(CompaniesExtractParams(domain="example.com"))`.
2. **Draft.** From the text write one `icp_prompt` sentence, two or three `phrase_match` candidates, a `category`, and the filters that are truly hard (country, business model). A candidate phrase has to distinguish the target, not merely appear on its site: if it would be strange for a company printing that phrase prominently _not_ to be in the ICP, keep it, otherwise drop it. Split hard filters from preferences: if the user would still contact a company without it, it is a preference, so leave it out of the filter and let ranking handle it. Target geography is a filter the user sets, not something to infer from where the seed customers happen to be based.
3. **Count each phrase** on its own. A phrase that counts in the hundreds is a filter; one in the tens is a seed list; one in the hundreds of thousands is noise.
4. **Sample 25.** Show domains and one-line descriptions. Ask the user to mark fits and misses in place.
5. **Correct and save.** Adjust from the marks, re-sample once, then save the query with a name and tags. Start every later job from that id.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 323kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 323kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 323kUpdated Claude CodeCodexVerifiedInstall