# Idea Generator > Generate business, marketing, and post ideas tied to your product's wedge and stage. Use when asked for partnership ideas, marketing tactics, content angles, or fresh experiments to run. - Skill: `codealive-ai/idea-generator` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add codealive-ai/idea-generator` - Raw SKILL.md: https://api.skillmd.com/api/skills/codealive-ai/idea-generator/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Marketing & Growth - Author: CodeAlive-AI (https://skillmd.com/u/codealive-ai) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/codealive-ai/idea-generator --- # Idea Generator Idea generation module for a CEO operating system. Job: produce **specific, actionable ideas** in three categories — partnerships, marketing tactics, social posts — all tied to the user's current wedge and stage. **Important:** before generating, read `MEMORY.md` for the current wedge, stage (pre-PMF / scaling / mature), differentiators, ICP, competitors, and recent priorities. Ideas should help the next learning step or proof point — not premature scaling. ## Reading Context Pull from MEMORY.md (or ask the user once if missing): - **Product** — one-sentence description, key differentiators - **Wedge** — the specific entry point you're attacking - **Stage** — pre-PMF / early-PMF / scaling - **Customers (ICP)** — role, company size, industry - **Competitors** — who shows up in deals - **Use cases** — top 3-5 jobs the product is hired for - **Distribution** — current channels that work If any of these are missing, ask the user once before generating, then save the answer back to MEMORY.md. ## Three Idea Categories ### A: Business Partnership Ideas Specific partnerships with named companies or company types. Think: - **Integration partnerships** — tools the customer already uses; embed your product into their workflow - **AI/agent partnerships** — agent platforms or assistants where your product can be the data/context provider - **Channel partners** — consultancies, system integrators, resellers who reach your enterprise ICP - **Co-marketing** — joint research, webinars, reports with adjacent vendors - **Open source** — contribute to standards/ecosystems where your buyers hang out - **Vertical** — partners specific to one industry where your wedge lands hardest - **Academia** — university programs, research partnerships in your domain ### B: Marketing Ideas Specific tactics with mechanics: - **Product-led growth** — free tier hooks, viral loops, referral programs - **Community** — Discord/Slack community, champions program, OSS contributors - **Content marketing** — SEO articles, comparisons, benchmarks, case studies - **Events** — relevant conferences, meetups, webinars (name them) - **Campaigns** — themed campaigns ("[Pain] Month", "[Workflow] Challenge") - **Influencers** — specific DevRel, YouTube, X/LinkedIn voices in the niche - **PR** — press releases, research reports, awards in the category ### C: Social Media Post Ideas With platform and format specified: - **LinkedIn** — thoughts for your buyer persona, industry observations, mini case studies - **X (Twitter)** — hot takes, polls, threads in your niche - **Dev.to / Hashnode / Medium** — technical deep dives, tutorials, "How [YOUR_PRODUCT] does X" - **HackerNews** — Show HN format, technical breakdowns - **YouTube** — demos, comparisons, "day in the life" content - **Reddit** — top subreddits where your ICP hangs out (name them) — valuable posts, not ads ## Validation Filters Run every idea through these silently. If an idea fails, fix it or drop it. ### Why Now test "Why is this idea possible/relevant NOW when it wasn't 6 months ago?" Every idea must connect to a recent shift — new technology, behavior change, market event, infrastructure change, regulatory shift. ### Tarpit detection Flag ideas that fall into crowded spaces where many have tried and failed. Generic tarpits to avoid: - "Generic AI [category] tool" — many competitors, commodity - "Another [popular IDE/platform] extension" — oversaturated channel - "General [job] productivity" — too vague, no wedge Off-the-beaten-path examples (good): - Underserved verticals where your wedge is uniquely strong - New categories that just opened (e.g. recent platform/standard adoption) - Adjacent jobs your product enables that competitors don't address ### Differentiation check "Does this idea leverage what your product **uniquely** has?" Tie back to differentiators from MEMORY.md. If any competitor could do this equally well — the idea is weak. ## Idea Format ``` [Category A/B/C] Short title What: what specifically to do (2-3 sentences) Why now: what changed that makes this timely (1 sentence) Why it works: tie to your differentiator (1-2 sentences) First step: one concrete action to start Risk: tarpit / crowded / safe (1 word) [link if found relevant] ``` ## Examples (template form) ### Example A (partnership): ``` [A] [Adjacent Tool] Integration What: propose [Adjacent Tool] (open-source [category], NK+ users) an integration where your product becomes their [data/context/workflow] provider. They actively seek better [missing capability]. Why now: [recent platform shift] is creating demand for this — [Adjacent Tool] just shipped support, perfect timing to plug in. Why it works: your product's [differentiator] delivers [N]x better [metric] than their built-in [alternative]. Win-win: they get a killer feature, you get distribution. First step: post in their community (#integrations channel) with a proposal + link to your integration spec. Risk: safe ``` ### Example B (marketing): ``` [B] "[Pain] Horror Stories → How We Fixed It" series What: 5 real (anonymized) cases where your product solved [pain]. Each case = blog + LinkedIn post + X thread. Why now: [pain] is the #1 unsolved problem in [adjacent narrative]. Nobody is doing case studies for it specifically. Why it works: horror stories go viral in your community. Educational + entertaining, not advertising. First step: ask current users for permission to anonymously describe their use case. Risk: safe ``` ### Example C (post): ``` [C] "I asked an AI to [do task] N times. Here's what it found." What: LinkedIn post with stats from your product — most common patterns, repeating mistakes, breakdown by category. Why now: [topic] is hot but nobody publishes real aggregate data. First-mover advantage on data-driven content. Why it works: your buyers love concrete numbers. Educational + product demo without direct advertising. First step: run your product on a representative dataset and collect stats. Risk: safe ``` ## Challenge Mode (optional) If the founder says "challenge me" or "challenge mode" — before generating ideas, ask 2-3 of these questions and use the answers to sharpen the output: - "What changed in the market this week/month that we should react to?" - "Which of our current ideas are we most excited about — and why might we be wrong?" - "What does our unique info diet tell us that other founders don't see?" - "If your product disappeared tomorrow, who would miss it most and why?" - "What's the one thing we know about our users that competitors don't?" Use the answers to filter and rank the generated ideas. ## How to Find Inspiration 1. Use **web_search** for fresh trends 2. Rotate queries (see HEARTBEAT.md if present) 3. Tie findings back to the wedge/differentiator from MEMORY.md 4. Do not fabricate facts or links 5. Check memory — do not repeat ideas already sent ## Tone - Informal, like talking to a fellow founder - With specifics (companies, numbers, mechanics) — pulled from research, not invented - No boilerplate phrases - Feel free to ask "What do you think?" or "Want me to dig deeper?" Follow output preferences from USER.md (language, format, platform constraints). ## Red Flags Check After generating ideas, append a brief "Red flags" block: ``` ⚠ Red flags check: - Tarpit risk: [any ideas that are in crowded spaces?] - No Why Now: [any ideas without clear timing driver?] - Trend-chasing: [any ideas driven by hype, not by a real problem?] - Copycat: [any ideas a competitor already does well?] ``` If all ideas pass — write "All clear, no red flags." ## When the Founder Writes Themselves Help them think in the context of their question. If they ask for more ideas in a category — generate. If they give feedback — log it to memory and incorporate next time.