gsj-platform-best-practices

name

gsj-platform-best-practices

description

Check and adapt GSJ ad and social copy against platform formatting rules and character limits for Google RSA, Meta (Facebook/Instagram), and LinkedIn. Use whenever GSJ copy is being written, reviewed, or fitted for one of these platforms — RSA headlines/descriptions/paths, Meta primary text/headline/description, LinkedIn posts/hashtags — or when someone asks whether copy "fits," is "within limits," or needs trimming for a platform. Counts characters exactly (never estimated) and, when copy is over, reports exactly how many characters over.

GSJ Platform Best Practices

Validate and adapt Groundswell (GSJ) ad/social copy against each platform's real formatting rules and character limits. The limits live in a spec file — always read them from the file, never from memory — and character counts are always exact, never estimated.

Source of truth

All limits and rules come from:

references/gsj-platform-specs.xlsx

Three sheets: Google RSA, Meta, LinkedIn. Do not hard-code or recall limits — if the file changes, the skill must follow it. To see the current specs:

python3 scripts/check_copy.py --show-specs
# or one platform:
python3 scripts/check_copy.py --show-specs --platform "Google RSA"

The exact-count rule (non-negotiable)

Never eyeball or estimate a character count. Every count must come from the script, which uses len() over the raw string. When copy is over a limit, state exactly how many characters over and how many to trim — the script prints this for you.

How to check a piece of copy

Run one field at a time. Prefer --stdin so quotes, apostrophes, emoji, and newlines in the copy can't break shell quoting:

printf '%s' "Cold-Pressed Organic Juice Delivered Daily" \
  | python3 scripts/check_copy.py --platform "Google RSA" --field "Headline" --stdin

Or pass it inline with --text "...". Platform names match the sheet tabs ("Google RSA", "Meta", "LinkedIn"); field names are matched case-insensitively and by prefix ("Headline", "Primary text", "Description", "Post length", ...).

The script prints the exact length, the rule, notes, and a verdict: OK — n/limit, x to spare or OVER by x characters — trim x. Exit code is 0 when within limit, 1 when over.

For a full ad or post, check each field separately (e.g. every RSA headline and description, each up to the file's stated maximum count of assets).

Workflow

1. Identify the platform and which field each piece of copy maps to.
2. Run check_copy.py per field (use --stdin). Report the exact count vs. limit for each.
3. For anything over, say by exactly how many characters, then rewrite to fit — preserving meaning and GSJ brand voice — and re-run the script to confirm the rewrite is within limit.
4. Apply the non-length rules from the file too (the Notes column and the fields with no numeric limit), e.g.:
   • Google RSA: provide enough headlines/descriptions per the file's min/max; each headline must stand alone; don't repeat the same claim; use pinning sparingly.
   • Meta: front-load the essential point in the first ~125 chars of primary text (mobile truncates hardest); supply the right image crop; don't rely on the description showing.
   • LinkedIn: keep external links out of the first line (put them in the first comment); prefer native video/carousels; write the first 2–3 lines to stand alone; end with a genuine question; 3–5 hashtags, no stuffing.
5. Fields with no numeric limit (image ratio, link display, pinning, native content, engagement window) are guidance — apply the rule, no count needed.

Notes on counting

• Counts are by character (Unicode code point), matching how these platforms count. An emoji counts as its code points.
• Meta and LinkedIn limits are engagement/truncation thresholds ("recommended", "no hard cap after ~N"); the script still reports exact overage against the number in the file so you can make an informed trim.

Brand voice and claims

This skill governs fit and formatting only. It does not check brand voice or factual/health/sustainability claims — run gsj-brand-voice and gsj-product-accuracy for those. Order: get the message right (voice + claims), then use this skill to fit it to each platform.