# ASTRO Website / Product Skill

Use this skill before editing Clintware public pages, product positioning, startup / YC pages, portfolio pages, navigation, or public visual/copy systems.

## Purpose

ASTRO exists to prevent generic AI-generated website output: oversized vague hero copy, repetitive card grids, generic SaaS claims, random metrics, repeated CTAs, fake polish, and roadmap language presented as current capability.

## Required sequence

`READ ASTRO -> INSPECT CURRENT PAGE/REPO -> IDENTIFY PAGE JOB -> PRESERVE WORKING PARTS -> REMOVE TEMPLATE NOISE -> WRITE FROM EVIDENCE -> SEPARATE NOW/NEXT -> QA -> DEPLOY -> VERIFY`

Before editing:

1. Inspect the current repository and actual page.
2. Preserve approved functionality, routes, forms, embeds, downloads, analytics, security controls, and accepted prior work.
3. Decide what job the page has for the visitor.
4. Write from evidence and product truth rather than marketing convention.
5. Separate current capability from testing, roadmap, or thesis.
6. Use compact, high-contrast, direct technical design.
7. Reject the draft if it could be reused for an unrelated AI startup by changing only the product name.

## Clintware visual / copy constraints

- Display brand: **Clintware™**.
- Display slogan: **GO FURTHEST.™**.
- No giant bubbly/loopy/generic SaaS hero typography.
- No forced `hero -> three cards -> feature grid -> repeated CTA` structure.
- Prefer product briefs, operator memos, evidence records, technical dossiers, professional profiles, or working interfaces depending on the page's actual purpose.
- Avoid hype-heavy adjectives and vague benefit stacks.
- Attach metrics to the role, company, decision, product, or operating change they prove.
- Keep one clear primary next action where possible.

## Shared component consistency

Shared navigation and control components must remain global rather than being restyled page by page.

- `clintware.js` owns the injected non-home route control (`Back / Home`).
- `typography-lock.css` owns its visual treatment.
- Do not add page-local `.cw-route-tools` styling unless a verified page-specific accessibility defect cannot be solved globally.
- Back/Home must remain clearly visible against the dark site background with readable text, explicit borders/backgrounds, keyboard focus states, and mobile-sized tap targets.
- When changing a shared component, test representative page types: custom hero/registry page, `.page-hero` page, utility/contact page, product/tools page, article/blog page, and individual skill page.
- Prefer one global fix over repeated local exceptions.  Remove obsolete page-local overrides after the global fix is verified.

## Page test

A visitor should quickly understand:

1. What is this?
2. Who is it for?
3. What problem/job does it address?
4. What exists or is proven today?
5. What is not yet shipped/proven?
6. Why should they care?
7. What is the sensible next action?

## Product truth order

Use:

`PROBLEM -> THESIS -> CURRENT PRODUCT -> EVIDENCE -> BOUNDARIES -> WHAT MUST BE PROVEN NEXT`

Avoid:

`TAGLINE -> BENEFIT CARDS -> GENERIC FEATURES -> REPEATED CTA`

## Quillgeist

Quillgeist is an **adaptive intent compiler for AI**, not merely a prompt optimizer, writing assistant, or generic copilot.

Preserve:

`TYPE -> UNDERSTAND -> IMPROVE -> EXECUTE -> LEARN`

- TYPE: user expresses normal intent.
- UNDERSTAND: resolve goal, context, constraints, tools, permissions, definition of done.
- IMPROVE: prepare the instruction/context required by the target model/tool.
- EXECUTE: run eligible deterministic work locally while reserving model reasoning for judgment.
- LEARN: use accepted/rejected edits, corrections, preferences, and outcomes to reduce repeated instruction.  This remains in development until proven.

## LandThePlane

Do not reduce LandThePlane to generated interview questions.

The durable asset is the candidate-owned evidence system: verified accomplishments, target-role requirements, answer structures, interview history, feedback, longitudinal improvement, post-hire accomplishment capture, and credible future pivots.

Interview preparation/review is the wedge because it provides urgency and feedback.  The company thesis depends on evidence reuse and repeat behavior across rounds and roles.

## Fall 2026 YC language

Quillgeist and LandThePlane may be described as **Fall 2026 Y Combinator application candidates** or **being prepared for a Fall 2026 YC application** while that is the actual plan.

Never imply acceptance, endorsement, affiliation, interview selection, or YC participation unless it actually occurs.

## Rejection checklist

Reject and rewrite if:

- the page could belong to any unrelated AI startup;
- the hero says almost nothing specific;
- marketing adjectives outweigh evidence;
- every section is a card grid because it was easy to build;
- roadmap looks shipped;
- caveats are hidden;
- the visitor must infer what the product actually does;
- the page looks like generic SaaS boilerplate instead of Clintware;
- a shared navigation/control component is restyled locally when the issue should be fixed globally;
- Back/Home is low-contrast, browser-default-looking, or visually inconsistent with other secondary pages.