Co-Messi/claude-skill-gain-exposure

Gain Exposure

Shipping is half the job. The other half is getting eyes on it. This skill fixes the three ways AI-driven launches fail:

1. Too narrow on platforms — one subreddit and stop. Not a launch.
2. Too narrow on media — text post, ignoring that short-form video often gets 10–100× the reach.
3. Bot-detected — DOM form fills → Reddit/LinkedIn/TikTok flag and shadowban.

Core rule — cast the widest possible net

The agent chooses scope. Not a fixed list. Before posting, look at the project and every platform where its audience lives, then go there. Don't stop at 3 subs if 10 are relevant. Don't skip video if the project has any visual surface.

Minimum coverage for a real launch:

• Reddit — as many relevant subs as the agent can identify (dynamically discovered, not fixed)
• Twitter/X thread with media
• Hacker News Show HN if it qualifies
• IndieHackers, Dev.to / Hashnode / Medium, Product Hunt (if polished), LinkedIn story post
• Discord / Slack communities user is in
• Short-form video — TikTok / YouTube Shorts / Instagram Reels
• Long-form video on user's YouTube
• Newsletters in the niche

If only one or two platforms have been hit, the task is not done.

High-level workflow

0. Intake — ask project + angle + audience + scope + credentials. See references/intake-and-preflight.md.
1. Launch the browser — see references/browser-setup.md (gstack-browse preferred, chrome-devtools-mcp fallback).
2. Auth check — verify logins on every target before touching anything. See references/intake-and-preflight.md.
3. Preflight — landing-page OG check, UTM registry, Reddit per-sub gate check. See references/intake-and-preflight.md.
4. Draft assets — per-platform titles + bodies, video pipeline if in scope (references/video-pipeline.md), images (references/image-generation.md).
5. Post everywhere — skip-on-fail, retry-once-at-end, escalate only twice-failed. See "Execution model" below.
6. Report — output table with every URL + timestamp + any deferred/skipped.

Execution model — autonomous with escalation

After intake, run end-to-end. Do NOT stop for confirmation between platforms.

Main loop:

1. Work the target list in order.
2. Target fails (auth expired, form bounced, 2FA, captcha, rate limit, upload rejected) → mark FAILED, log reason, immediately continue.
3. After all targets attempted once → loop back and retry every FAILED once. Fresh run: reload, wait longer, re-screenshot.
4. Still FAILED after retry → escalate to user with failure screenshot + suggested next action.

Post-and-forget. Job ends when posting is done. Don't monitor engagement or reply to comments — not in scope. Report the URL table and end.

Detailed references

Topic	File
Browser setup (gstack-browse + chrome-devtools-mcp fallback, cookie import, tab hygiene)	browser-setup.md
Step 0 intake + Step 1 auth check + Step 2 preflight (OG / UTM / Reddit sub rules)	intake-and-preflight.md
Anti-detection: visual reasoning, timing, blocked handling, ban-avoidance rules	anti-detection.md
Image gen via ChatGPT / Gemini web UI	image-generation.md
Video pipeline (Remotion sizzle + ElevenLabs TTS + clipping tools)	video-pipeline.md
Per-project community targeting + per-platform writing	community-and-writing.md
18 additional high-leverage tactics	tactics.md
Cost efficiency (Haiku + Playwright runner for 5+ sub launches)	cost-efficiency.md

Read the relevant reference file before executing that phase. Don't try to run the whole skill from this summary alone.

Output format (end of run)

Launch report — <project name>

Reddit:
  - r/<sub1>: <url> — <time>
  - r/<sub2>: <url> — <time>
Twitter thread: <url>
Hacker News: <url>
IndieHackers: <url>
Dev.to: <url>
LinkedIn: <url>
TikTok: <url>
YouTube Shorts: <url>
Instagram Reels: <url>
YouTube long-form: <url>
Discord/Slack posts: <server> — <channel>
Deferred: <platform> (<reason>)
Skipped: <platform> (<reason>)