TeqfocusAdmin Teqfocus Post Scorer
Predicts post performance by scoring a draft against what has actually worked on Teqfocus company and personal profiles — not against generic LinkedIn advice.
Prerequisite: Read /mnt/skills/user/teqfocus-gtm/SKILL.md first. This skill scores performance; linkedin-post-reviewer scores messaging alignment. A post must pass both.
Data sources (in priority order)
- LinkedIn analytics export (xlsx/csv) uploaded by the user — impressions, engagement rate, CTR per post
- Apify LinkedIn scraper if
APIFY_API_TOKENis set and the user approves the pull - Manual history: user pastes 5–10 recent posts with rough performance notes
- If none available: score against the heuristic rubric below and flag the score as
[UNCALIBRATED — no history provided]
Build the performance baseline
From history, extract per-post: hook type (question / stat / contrarian / story), length, format (text / carousel / image / video), topic pillar, CTA type, posting persona (company page vs. personal). Compute median engagement per attribute. Identify the top-quartile pattern — this becomes the target profile.
Scoring rubric (100 points)
| Dimension | Points | What earns them |
|---|---|---|
| Hook strength | 25 | First 2 lines stop the scroll; matches a hook type in the account's top quartile |
| Pattern match | 20 | Format + length + topic pillar match historical top performers |
| Persona relevance | 20 | Speaks to a named Teqfocus persona's real problem (CIO/CDO/RevOps/VP Sales) |
| Readability | 15 | Line breaks every 1–2 sentences, no wall of text, mobile-scannable |
| CTA clarity | 10 | Exactly one CTA, specific and frictionless |
| Voice compliance | 10 | Zero banned words, no banned openers, senior-practitioner tone |
Verdict bands
- 80–100 — Ship it. Minor polish only.
- 60–79 — Rework the flagged dimension. Provide the specific rewrite, not just the critique.
- Below 60 — Rebuild. Route back to linkedin-campaigns with the diagnosis attached.
Output format
Score table → verdict → the 1–3 highest-impact rewrites shown as before/after → predicted performance band relative to account median (e.g., "top quartile likely" — never invent impression numbers, flag [VERIFY] on any quantitative claim).
Rules
- Never fabricate historical performance data. If the baseline is thin, say so.
- A high score never overrides a linkedin-post-reviewer fail — messaging alignment wins.
- Scored drafts still pass through humanizer before delivery.