By name: summarize-100 description: Compress any topic, document, answer, profile, or message into approximately 100 words — the 100 most semantically-dense tokens that preserve the core signal. Use when the user asks for "100 palabras", "100 words version", "versión 100", "tldr 100", "elevator pitch", "comprime a 100", "resumen 100", or any equivalent request for a token-budgeted summary. The goal is maximum information density per word — every token must earn its place.
summarize-100 — 100-Word Compression Skill
Purpose
Compress arbitrary input (a document, a topic, a previous answer, a CV, an About section, a meeting note) into approximately 100 words that carry the most semantic weight. Optimized for:
- LinkedIn About sections (mobile-truncated)
- Executive summaries / TL;DRs
- Elevator pitches
- Headline / hero copy
- Token-budgeted prompt contexts
- Twitter / X long-form posts
- Bio paragraphs
Triggers
The skill activates when the user requests a token-budgeted summary near 100 words:
- Spanish: "100 palabras", "versión 100", "comprime a 100", "resumen 100", "elevator pitch", "en cien palabras"
- English: "100 words", "100-word version", "tldr 100", "tldr in 100", "compress to 100", "elevator"
- Numeric variants: "~100", "around 100", "unas 100"
Workflow
- Identify the thesis (1 sentence) — if everything else were lost, what single claim must survive?
- Find 3-5 pillars — concrete supporting facts, numbers, proper nouns, deliverables. Drop generics.
- State the outcome / call-to-action — what should the reader do, conclude, or remember?
- Draft to ~120 words first — overshoot, then cut.
- Cut ruthlessly (see anti-patterns below).
- Verify count — count words explicitly. Target: 90-110 words. Report the count.
Best Practices
- Lead with the noun or verb that carries weight — "AI Engineer, 20 years..." not "I am an AI Engineer who has 20 years...".
- Use hard numbers over vague modifiers — "20+ years" beats "extensive experience". "$10M ARR" beats "significant revenue".
- Prefer proper nouns over categories — "Snowflake, Postgres, Azure" carries more information per token than "modern data warehouses".
- Group related items with commas — "Python, C#, TypeScript" packs 3 concepts in 4 tokens.
- Em-dashes (—) compress two clauses into one without losing flow.
- Keep one signature line — a memorable claim ("imagination is the only ceiling", "remote-native for years"). It anchors recall.
- Concrete > abstract — "shipped 12 RAG chatbots" beats "delivered AI solutions".
- Active voice always — "Built X" beats "X was built by".
Anti-Patterns (Cut on Sight)
| Cut this | Replace with |
|---|---|
| "in order to" | "to" |
| "due to the fact that" | "because" |
| "is responsible for X" | the verb of X |
| "is involved in" | the verb of the involvement |
| "various / several / many / numerous" | a number, or nothing |
| "innovative / robust / cutting-edge / world-class" | a concrete proof |
| "passionate about" | the action that proves it |
| "extensive / comprehensive / broad experience" | years + a stack |
| Hedges: "perhaps / arguably / somewhat / kind of" | delete |
| "As a [role], I..." opener | start with the verb or claim |
Output Format
- 3-4 short paragraphs, plain prose.
- No bullet lists inside the 100-word body (lists fragment the count and waste structural tokens). Lists are OK outside the body if context demands.
- Always end with the word count, e.g.
— 94 words. - If the result lands outside 90-110, redo before delivering.
Token Density Heuristic
Each word should answer at least one of:
- Who? (proper noun, role, identity)
- What? (concrete deliverable, technology, artifact)
- How much? (number, scale, duration)
- Why does it matter? (outcome, differentiator)
- What next? (CTA, availability, status)
If a word answers none of these — cut it.
Examples (Patterns, Not Templates)
LinkedIn About (English, professional)
Senior AI + Data Engineer, 20+ years shipping production systems — ETL, Snowflake, RAG chatbots, trading engines. Author of [Project] ([link]), open-source [domain] framework. Stack: Python, C#, TypeScript, Azure. Remote-native across US, EU, LATAM. Based in [city], available [timezone]. — 42 words.
Topic explainer (Spanish, technical)
Tema X resuelve el problema Y mediante Z. Tres pilares: A, B, C. Ventaja clave: métrica concreta. Limitación: caso conocido. Útil cuando [contexto]; evítalo cuando [contraindicación]. — N palabras.
Lessons Learned
(append patterns here as they emerge across uses)
- 2026-05-16 — Created from LinkedIn About compression request. Pattern: when compressing a profile, always preserve (1) seniority signal, (2) one flagship project name, (3) stack list, (4) location/availability. Cut adjectives first, then connectors, last the verbs.