Consulting LinkedIn Hooks
Workspace: use the selected project and its AGENTS.md; keep existing entity folders and
Reality headings. Business paths below are relative to that project. Bundled resources are relative
to this installed skill; sibling capabilities resolve by their installed names. Never search another
private checkout for missing inputs. Use workspace identity, audience, pricing, and DESIGN.md fonts.
Local _work adapters and evals are optional workspace tools, not bundled dependencies. Check
presence and current help first; otherwise use an available connector for the same scoped operation.
If neither exists, report that step incomplete. For missing scorers, perform the stated checks and
label the result manual/unscored; never invent a numeric score or successful provider action.
Line 1 is ~80% of a LinkedIn post's performance — it decides whether anyone reads line 2. This skill
generates and critiques hooks from patterns measured against real engagement, not generic advice.
Patterns + cited examples + the empirical rules live in references/hook-patterns.md (references/hook-patterns.md) — read it before writing or grading a hook.
The one rule
Steal structure, not persona or slop. The patterns come from creators (Justin Welsh, Jasmin Alić,
Ruben Hassid, …) whose voice the owner doesn't share. Map the shape onto the owner's substance (AI/agents for
music & media; B2B AI adoption) and run the result through recoup-internal-consulting-copy-writer (no banned words,
no em-dash). Ruben Hassid's AI hooks are the closest direct transfer — start there for AI content.
To WRITE hooks
- Name the post's job and the engagement goal. Comments, or saves/shares? That picks the pattern family: belief/identity → comments; utility/interrupt → saves/shares (see the engagement-signature note per pattern in the reference).
- Generate 8–12 options across at least 4 different patterns (don't give 10 variants of one). Pull
from the 8 patterns in
references/hook-patterns.md. Keep each ≤60 characters, one idea, no question, no listicle count (a vivid specific number is fine; a count is not). - Write the one-two punch: for the top 3, draft line 2 as well (a ≤8-word amplifier/twist) — the hook is really the first two lines before the fold.
- Run voice + anti-slop: pass the finalists through
recoup-internal-consulting-copy-writer. Cut any banned word, em-dash, or borrowed-persona edge. - Hand off: for a full post around the hook, continue in
recoup-internal-consulting-linkedin-post-architect.
To CRITIQUE a hook
Grade it against the reference's cross-cutting rules and name the failing one specifically:
- Over 60 chars / two ideas / buried lede → too long; cut to the claim.
- A question, or a listicle count ("5 ways…") → swap for a declarative pattern.
- Vague/clever instead of concrete → rewrite with a specific.
- Doesn't match the engagement goal → switch pattern family. Then rewrite it 2–3 ways across different patterns and say which you'd ship and why.
Keep it current
Hooks decay; "best practice" is just "what got engagement last month." Re-run the scrape +
swipe/_work/analyze_hooks.py monthly (or via recoup-internal-consulting-integrations-sync) and refresh
references/hook-patterns.md from the new top set. Origin evidence + method:
swipe/2026-06-25-hook-analysis.md.