content-engine

name	content-engine
description	Create platform-native content systems for X, LinkedIn, TikTok, YouTube, newsletters, and repurposed multi-platform campaigns. Use when the user wants social posts, threads, scripts, content calendars, or one source asset adapted cleanly across platforms. Do NOT use for deriving the voice profile itself (brand-voice), for actually publishing/distributing the posts (crosspost), or for long-form articles (article-writing).
origin	ECC

Content Engine

Turn one idea into strong, platform-native content instead of posting the same thing everywhere.

When to use

• writing X posts or threads
• drafting LinkedIn posts or launch updates
• scripting short-form video or YouTube explainers
• repurposing articles, podcasts, demos, or docs into social content
• building a lightweight content plan around a launch, milestone, or theme

When NOT to use:

• the ask is distribution mechanics (staggering, per-platform posting APIs) — that is crosspost
• the ask is long-form (guide, essay, newsletter issue) — that is article-writing
• no source asset or idea exists yet — get the anchor first; content generated from nothing is filler by construction

Inputs

Required before drafting — absent input = stop and ask, never proceed on assumptions:

1. Source asset — what we are adapting from (article, video, demo, memo, launch doc, or a single stated idea). No anchor → no drafts.
2. Audience — builders, investors, customers, operators, or general.
3. Platform targets — X, LinkedIn, TikTok, YouTube, newsletter, or multi-platform.
4. Goal — awareness, conversion, recruiting, authority, launch support, or engagement.
5. Voice profile — from brand-voice if one exists. For SkyyRose work, product facts come only from wordpress-theme/skyyrose-flagship/data/skyyrose-catalog.csv and collection language only from docs/brand/collection-stories.md — never from memory.

Procedure

1. Confirm the four inputs above.
2. Extract 3–7 atomic ideas from the anchor asset. One clear idea per post.
3. Draft platform-native variants — never the same copy resized:
   • X — open fast; one idea per post or per tweet in a thread; links out of the main body unless necessary; no hashtag spam.
   • LinkedIn — strong first line (visible before "see more"); short paragraphs; explicit framing around lessons, results, and takeaways.
   • TikTok / short video — first 3 seconds interrupt attention; script around visuals, not just narration; one demo, one claim, one CTA.
   • YouTube — show the result early; structure by chapter; refresh the visual every 20–30 seconds.
   • Newsletter — one clear lens, not a bundle of unrelated items; skimmable section titles; the opening paragraph does real work.
4. Trim repetition across outputs and align each CTA with platform intent — small, clear asks.
5. Run the Verification checks on the draft files before delivering.
6. When asked for a campaign, return: the core angle, platform-specific drafts, optional posting order, optional CTA variants, and any missing inputs needed before publishing.

Verification

Write each platform draft to its own file (e.g. x.txt, linkedin.txt) so the checks have a real target. A check that errors (missing file → grep exit 2) is a dead gate, not a pass (bug-230) — fix the path and re-run.

1. Hard length limits — platforms truncate or reject silently:

awk 'length($0) > 280 {print "OVER-LIMIT line " NR ": " length($0) " chars"; bad=1} END {exit bad}' x.txt

PASS: no output, exit 0. Any OVER-LIMIT line = trim before delivery. Adjust the limit per platform (X 280 · Threads 500 · Bluesky 300 · LinkedIn 3000). [repro]

2. No cross-platform copy duplication — the core rule, made falsifiable:

sort x.txt linkedin.txt | uniq -d | grep -v '^[[:space:]]*$'

PASS: no output, exit 1 from the final grep. Any surviving line appears verbatim on two platforms — rewrite one of them. [repro]

3. Banned-trope scan — same gate as brand-voice:

grep -nEi 'excited to share|no fluff|game.chang|cutting.edge|revolutionary' x.txt linkedin.txt

PASS: exit 1 (zero matches). [repro]

Both check 1 and check 2 were observed failing on deliberately broken input before being trusted (rule 3 of the authoring standard) — see the worked example.

Worked example

Real run, 2026-07-28. Anchor: a Black Rose one-run drop announcement. Drafts written to the session scratchpad as x.txt and linkedin.txt.

First X draft came out at 293 chars — the length gate went red, proving it can fail:

$ awk 'length($0) > 280 {print "OVER-LIMIT line " NR ": " length($0) " chars"}' x.txt
OVER-LIMIT line 1: 293 chars

After trimming (wc -c → 238):

$ awk 'length($0) > 280 {print "OVER-LIMIT line " NR ": " length($0) " chars"; bad=1} END {exit bad}' x.txt && echo "PASS: all lines <= 280"
PASS: all lines <= 280

$ sort x.txt linkedin.txt | uniq -d | grep -v '^[[:space:]]*$'; echo "exit=$?"
exit=1

All outputs observed this session [repro]. The X version opens "Black Rose is armor." — the LinkedIn version reframes the same fact as a make-story with explicit takeaway; zero shared lines.

Failure modes

Failure	What it looks like	Rule
Cross-posting identical copy	uniq -d returns shared lines	Never ship it; each platform gets a native adaptation — that is this skill's core contract
Hallucinated product facts	price, SKU, or drop date stated from memory	bug-096 shape: plausible ≠ sourced. Catalog CSV is the only product manifest; data/sot-images.json the only imagery source
Mixed collection canon	"bloodline" copy on a Signature post	Each collection has its own locked canon (docs/brand/collection-stories.md); never mix quotes
Length overflow shipped	platform truncates the CTA off the end	Run check 1 per file, per platform limit — silence from an unrun gate is not a pass (a SKIP is not a PASS; the drafter closes it before handoff to crosspost)
Hooks that summarize	first line restates the title	Hooks matter more than summaries — open with the sharpest specific, not the topic
Generic hype language	"game-changing", "excited to share"	Check 3 catches these mechanically; rewrite with specifics over slogans