content-atomizer — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited content-atomizer (Agent Skill) and scored it 96/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 1 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 1 flagged
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Break one substantial piece of content into many platform-native pieces. Prime directive: adapt, never excerpt — each atomic piece must stand alone as if written for that platform on purpose, carrying the user's voice, while format and length obey the platform. The second rule: atomize ideas, not paragraphs — you extract the content's core ideas first, then re-express each idea per platform.
skills/marketing/seo-content — then come back.skills/marketing/tweet-writer — it does niche research this skill skips.skills/marketing/newsletter.Ask in one batch, only what's missing:
skills/marketing/brand-voice output if it exists).Don't ask which formats they want — that's your job to propose in Step 3. If platforms are unknown, default to LinkedIn + X + newsletter item and say so. State assumptions and proceed — don't stall.
Read the seed content and pull 5–10 discrete ideas, each tagged by type — the type determines which formats it can power:
If you can't find at least 4 distinct ideas, the content is too thin for a full pack — tell the user and scale scope down.
Pick 5–7 formats, never all 15 — quality over coverage (heuristic: a strong pack on the user's 2 priority platforms beats a thin spray across 8). Decision rules:
Format spec table (length and structure per format):
| Format | Yield from one seed | Spec |
|---|---|---|
| LinkedIn posts | 3–5 | 100–200 words; hook line → story/insight → lesson → question or CTA; line breaks for scannability |
| X/Twitter thread | 1–2 | 5–10 tweets; hook tweet ≤250 chars; one idea per tweet; closer with CTA (full craft: skills/marketing/tweet-writer) |
| Single tweets | 3–5 | ≤110 chars preferred; sharpest standalone insights |
| Newsletter item | 1 | 300–500 words; hook → insight → lesson → link to full piece |
| Short-video scripts (TikTok/Reels/Shorts) | 2–3 | 30–60s; hook (≤3s) → problem → insight → CTA; write as spoken lines with [B-ROLL] notes |
| Carousel outline | 1–2 | 5–10 slides; slide 1 = hook, one point per slide, last slide = CTA; text per slide ≤25 words |
| Quote graphics (specs) | 3–5 | One pulled quote + attribution; supply exact text, no design |
| Instagram captions | 2–3 | 100–150 words; hook first line (feed truncates), story → lesson → CTA |
| Slide deck outline | 1 | 10–15 slides: title → problem → 3–5 content → conclusion → CTA |
| FAQ/explainer | 1 | 5–10 Q&As, 40–80 word answers (feeds SEO FAQ blocks) |
| Case study | 1 | 1,000–1,500 words; situation → action → result → lesson (only if seed contains a real result) |
| Webinar/talk outline | 1 | Timed outline: intro → problem → solution → examples → Q&A → CTA (only on request) |
Show the user a one-screen plan first: idea inventory + chosen formats + which idea powers which piece. Get a quick confirm (or proceed if they said "just do it"). Then write every piece in full — no "you could write a post about X" stubs. While generating:
Order the pieces for posting: lead with the strongest hook on the priority platform, don't post two derivatives of the same idea on the same platform back-to-back, thread/carousel mid-pack when early pieces have tested the angle. Present as a numbered posting order (the user maps it to their own dates/scheduler — no calendar prescriptions).
# Atomization Pack — [seed content title]
## Idea inventory
| # | Idea (one line) | Type | Powers |
| 1 | ... | framework | Carousel, LinkedIn #2 |
(5–10 rows)
## Pack plan
Formats chosen: [list] — [one-line rationale tied to platforms/goal]
Skipped: [notable exclusions + why]
## Pieces
### LinkedIn Post 1 — [idea #]
[Full post text, ready to paste]
CTA: [the one ask]
### X Thread — [idea #s]
1/ [tweet]
2/ [tweet]
...
(char count noted per tweet)
### [Every other piece, fully written, grouped by platform]
## Posting order
1. [Piece] — [platform] — [why first]
2. ...
(strongest hook first; same-idea derivatives spaced apart)
## Leftovers
[Ideas from the inventory not yet used — future seed material]Check every piece before delivering:
skills/marketing/seo-content — primary source of seed articles; atomize each published article as a standard follow-up.skills/marketing/brand-voice — supplies the voice guide every piece must pass.skills/marketing/tweet-writer — takes over when the X pieces deserve dedicated niche research and hook testing; hand it the idea inventory.skills/marketing/newsletter — consumes the newsletter item; strong issues flow back here as seeds.skills/marketing/email-sequences — value emails can be atomized into social pieces and vice versa.~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.