产品目录站提交包

用 directory-launch-kit 技能帮我的产品准备目录站提交文案和图片，然后告诉我先提交哪几个。

directory-launch-kit

Goal: the user gets a short, honest list of directories that fit their product, ready-to-paste copy for each form, correctly sized images, and a tracker, and they do every submission and payment. This skill never creates accounts, submits forms, pays, upvotes or posts on the user's behalf.

Step 1 — Read the product, don't invent it

Collect, from the repo and the live site (ask only for what is missing):

Field	Where to find it
name, url (live https://, own domain)	site metadata, package.json, README
tagline (≤ 60 chars) and description (≤ 500)	the home page hero and meta description
pricing	the pricing page, exactly as written
ai / open_source / developer_tool / agent_skills	what the product actually is
competitors	ask the user (SaaSHub needs them)
maker_comment	the user's own words: draft it with them, never post it for them
logo + 2–5 real screenshots	public/, or capture the live site (Step 3)

Write it to product.json using assets/product.example.json as the shape. Copy rules are in references/copy-guide.md. Never add numbers, testimonials, user counts or "#1" claims the user can't prove; directories reject them and readers stop trusting the page.

Step 2 — Re-check the directory data

python3 scripts/check_directories.py            # OK / CHANGED / BLOCKED / DOWN per directory

references/directories.json records each directory's submit URL, observed cost, rules and the date it was read. Before you quote any price, open that directory's page (browser if BLOCKED or CHANGED) and say the date you read it. Update the JSON when something changed.

Step 3 — Real screenshots

Use screenshots of the real product, not mockups. If the site is live, capture it:

npx playwright screenshot --viewport-size=1440,900 https://example.com shot-1.png
npx playwright screenshot --viewport-size=1440,900 https://example.com/pricing shot-2.png

(or any browser at 1440×900). Show the product doing its job: the main screen, a result, pricing. Sizes per platform: references/assets.md.

Step 4 — Build the kit

python3 scripts/make_kit.py --product product.json --out launch-kit

Writes:

• launch-kit/copy.md: one section per directory: submit URL, observed cost and date, rules, each form field with n/limit, the maker comment. Fields over a limit are flagged OVER LIMIT and the script exits 1: shorten the text in product.json with the user and re-run. It never truncates.
• launch-kit/tracker.csv: id, directory, submit_url, cost_observed, checked, backlink_required, status, submitted_on, listing_url, link_rel, notes. Re-running keeps statuses already recorded.
• launch-kit/images/: ph-thumbnail-240.png, logo-512.png, og-1200x630.png, gallery-N-1270x760.png, screenshot-N-1920x1080.png, letterboxed (never cropped), each < 3 MB.

Directories the product doesn't fit are skipped with the reason (AI-only directories for non-AI products, OpenAlternative for closed source, DevHunt for non-dev tools, agent-skill marketplaces unless agent_skills: true). Show HN gets a warning if trying the product needs a signup.

Step 5 — Order the submissions (no dates, just order)

1. Free directories with long review queues first (Fazier free, Dang.ai free, SaaSHub, Launching Next, Uneed free): their queues run while you finish the rest.
2. Communities where the user must be present to answer (Show HN, Indie Hackers, Reddit, V2EX): one at a time, only when the user can reply to comments for the next hours.
3. Product Hunt: a draft first, launch when the user can stay on it all day.
4. Paid upgrades only if the user decides the price is worth it; list them with the observed price and date and let the user choose. Never pay or enter a card.

Directories that require a backlink (marked in copy.md): tell the user exactly which badge/link the directory wants and where it will appear on their site, and add it only if they agree.

Step 6 — Track what went live

After the user submits, they paste each listing URL into tracker.csv. Then:

python3 scripts/check_listing.py --tracker launch-kit/tracker.csv --domain example.com --write

marks rows live (with follow / nofollow), no-link or error.

Rules

• Never submit, sign up, pay, upvote, comment or post for the user. Draft; they act.
• BetaList offers an official agent API that only drafts (betalist.com/agent): you may create the draft only if the user asks, then hand them the claim link; they review, pay and publish.
• SaaSHub explicitly asks AI agents to let the human submit: don't automate it.
• No vote rings, no "please upvote" messages, no fake accounts. Product Hunt and HN penalise them.
• Quote prices only as "observed on <date> at <url>".

Files

• scripts/make_kit.py: product.json → copy.md, tracker.csv, images/
• scripts/check_directories.py: live re-check of every directory page and recorded price
• scripts/check_listing.py: confirm listings link to you, follow vs nofollow
• scripts/test_kit.py: self-test (offline)
• references/directories.json: the data (26 directories, checked 2026-11-03)
• references/directories.md: the same as a readable table
• references/copy-guide.md: taglines, descriptions, Show HN titles, the maker comment
• references/assets.md: image sizes per platform and how to capture them

Done when

• product.json contains only facts from the product and the user's own words
• check_directories.py was run and CHANGED/BLOCKED rows were read by hand
• make_kit.py exits 0 (no OVER LIMIT fields) and the images were opened and look right
• The user has copy.md, tracker.csv and images/, and knows the order to submit in