# Startup Channel Finder

> Find, verify, rank, and plan the best current channels to launch a startup using public evidence about audience fit, problem relevance, channel activity, promotion rules, timing, and effort. Use when Codex needs to decide where to launch a startup, SaaS, app, developer tool, open-source project, waitlist, beta, or new feature; research relevant communities, forums, directories, newsletters, marketplaces, and launch platforms; prepare channel-specific launch drafts; or create a sourced seven-day distribution plan without posting automatically.

- Skill: `kappaemme-git/startup-channel-finder` (Agent Skill, multi-file: 6 files)
- Install (CLI): `npx skillmds@latest add kappaemme-git/startup-channel-finder`
- Raw SKILL.md: https://api.skillmd.com/api/skills/kappaemme-git/startup-channel-finder/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: kappaemme-git (https://skillmd.com/u/kappaemme-git)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/kappaemme-git/startup-channel-finder

---


# Startup Channel Finder

Turn a startup URL or product description into an evidence-backed distribution plan. Recommend places where the intended audience is demonstrably active, not generic lists of popular launch sites.

Read [references/research-playbook.md](references/research-playbook.md) before researching. Read [references/analysis-schema.md](references/analysis-schema.md) before writing the analysis JSON.

Resolve `SKILL_DIR` as the directory containing this `SKILL.md`. Run bundled scripts from `SKILL_DIR`, while writing all outputs into the user's current workspace.

## Workflow

### 1. Understand the launch

- Inspect the supplied URL, repository, landing page, README, or product description.
- Identify product, primary ICP, painful job, differentiation, stage, geography, language, CTA, and launch goal.
- Separate facts from inferences. Ask one concise question only when ambiguity would materially change the audience or market; otherwise continue with labeled assumptions.
- Treat a feature launch differently from a company launch, waitlist, beta, open-source release, or paid-product launch.

### 2. Build a channel hypothesis map

- Translate the ICP and painful job into search phrases, problem language, role names, alternatives, workflows, and buying or adoption triggers.
- Search across several relevant channel types: communities, forums, directories, launch platforms, newsletters, marketplaces, social discussions, events, and niche publications.
- Use current web research. Never rely on a bundled static list as proof that a channel is active or appropriate today.
- Prefer original channel pages, official submission pages, official rules, and recent discussions over search-result snippets or third-party listicles.

### 3. Verify every candidate

For each candidate, record:

- direct channel URL and channel type
- evidence that the ICP is present
- evidence that the target problem or workflow appears there
- recent activity date when visible
- official rules or submission instructions
- promotion policy: `allowed`, `conditional`, `prohibited`, or `unknown`
- best format, angle, CTA, timing, and principal risk
- date on which the evidence was checked

Do not recommend a channel as actionable without direct audience evidence and a verified promotion or submission path. Mark unavailable facts as unknown; never invent audience size, reach, engagement, pricing, acceptance rate, or conversion potential.

### 4. Score deterministically

Score each factor from 0 to 5:

- `audience_fit` — How closely the active audience matches the primary ICP.
- `pain_presence` — How clearly the target problem or job appears in current discussions or channel positioning.
- `permission_fit` — How compatible the launch is with official rules and accepted formats.
- `timing` — How active and timely the channel is for this launch now.
- `effort_efficiency` — Expected value relative to submission, participation, or production effort.

Write unscored research to `outputs/startup-channel-analysis.json`, then run:

```bash
python3 "$SKILL_DIR/scripts/score_channels.py" \
  outputs/startup-channel-analysis.json \
  outputs/startup-channel-scored.json
```

The script computes the weighted score, verification status, and recommended action. Do not manually override its result. Explain any research limitation rather than inflating a factor.

### 5. Use action labels precisely

- `POST_NOW` — Strong fit, current evidence, and explicit permission for the intended format.
- `CONTRIBUTE_FIRST` — Strong audience fit, but community participation or trust should precede promotion.
- `SUBMIT` — A verified directory, marketplace, newsletter, publication, or launch platform with a suitable submission path.
- `SKIP` — Weak fit, inactivity, prohibited promotion, unknown rules, insufficient evidence, or poor effort-to-value ratio.

Never convert `SKIP` into a recommendation merely to fill the report. Prefer three defensible channels over twenty generic names.

### 6. Draft for the channel

- Create a distinct draft for each actionable top channel.
- Match the accepted format, tone, title pattern, disclosure expectations, link policy, and CTA.
- Ground the angle in the audience's job or problem; do not paste the landing-page headline everywhere.
- Avoid fake personal familiarity, fabricated traction, engagement bait, mass posting, ban evasion, or pretending to be a satisfied user.
- Keep drafts as suggestions. Do not post, submit, join, message, comment, or create accounts without a separate explicit request and authorization.

### 7. Build the launch sequence

- Order actions by dependency and learning value, not only score.
- Use early conversations and community feedback to improve later directory or platform launches.
- Create a seven-day plan with one goal, concrete actions, required asset, and success signal per day.
- Recommend manual tracking with channel, date, message variant, visits, signups, replies, qualified conversations, and relevant conversion event.
- Keep expected outcomes as hypotheses unless the user supplies historical data.

### 8. Generate the native report

Create:

- `outputs/startup-channel-scored.json` — machine-readable evidence, scores, and actions.
- `outputs/startup-channel-report.md` — user-facing report that opens directly in Codex and renders on GitHub.
- `outputs/startup-channel-calendar.csv` — execution sheet for the seven-day plan.

Run:

```bash
python3 "$SKILL_DIR/scripts/generate_report.py" \
  outputs/startup-channel-scored.json \
  outputs/startup-channel-report.md \
  --csv outputs/startup-channel-calendar.csv
```

Verify that every material claim and recommendation links to direct evidence. Return a clickable absolute link to the Markdown report.

## Report order

1. Launch verdict and strongest channel
2. Product, ICP, goal, and assumptions
3. Ranked channel table
4. Detailed playbook for each actionable channel
5. Seven-day launch calendar
6. Tracking plan
7. Skipped channels and reasons
8. Evidence limitations and unknowns

## Modes

- `quick` — Research up to five candidates and draft for the strongest two.
- `standard` — Research eight to twelve candidates across several types and build the full plan.
- `deep` — Research up to twenty candidates with broader source diversity and contradiction checks.
- `community-first` — Prioritize communities and trust-building before promotion.
- `directories-only` — Prioritize verified submission platforms and directories.
- `launch-week` — Prioritize an immediate seven-day sequence around a known launch date.

Use `standard` by default.

## Quality bar

- Link the product interpretation and every material channel claim to current public evidence.
- Require direct audience evidence and a rules or submission source for actionable channels.
- Make check dates, uncertainty, stale evidence, and contradictory signals visible.
- Distinguish audience presence from proven conversion or willingness to pay.
- Do not treat follower counts, member counts, or generic popularity as audience fit.
- Do not recommend prohibited promotion, private groups without access, scraped contact lists, or automated cross-posting.
- End with a sequence the founder can execute manually, not merely a list of websites.