SuperAPI AI Directory

run-organic-social-account-audit

Marketing & growth

View repository

Updated 10-27 11:24

About this skill

Account audit

Run before the first calendar, and lightly (reconciliation only) on later runs.

1. Which accounts exist and what each is for

  • gooseworks:brand_read sections: ["accounts","sources","organic"]. organic.own_accounts lists the brand's own TikTok/IG handles. sources lists saved public sources, including relationship: self social accounts. accounts lists connected accounts.
  • gooseworks:social_account_list: the TikTok/IG accounts the product manages, with their setup status. Read each with gooseworks:social_strategy_get (pillars, voice, rules, plug-ratio cap).
  • The brain has no record of LinkedIn, X, Threads, YouTube, Pinterest or Facebook accounts today, so check social/accounts.md in the workspace for a previous run's answers. Ask only for what is still missing: handles, which ones matter, brand vs founder, who approves founder posts, the time zone, the tool they post from, their free windows for the first hour after posting.

2. What has worked (per account)

  • TikTok/IG: own_posts from brand_read organic give caption, posted_at, views, likes and comments. gooseworks:social_analytics_get adds more for managed, connected accounts (it says not_available otherwise).
  • More history, or LinkedIn, X and Threads: read the account's own posts through the proxy (the marketing-data-access skill (../marketing-data-access/SKILL.md); verified endpoints only, as listed in ../run-organic-social/references/voice-card.md). Use the author's own posts only. For X: /v1/twitter/profile, then /v1/twitter/user-tweets (popular posts, not the newest: check created_at; the numbers are in legacy). This runs under the spend-gate skill (../spend-gate/SKILL.md) branch B: one yes to a ceiling, asked in the draft-report message (SKILL.md step 2), with the probe run inside it and every charge logged. Without the yes: the public profile in the host browser, or 5–10 pasted posts.
  • State the window you actually got (e.g. "7 IG posts, Oct 5–12"). The server returns recent posts only, and some captions come back cut off mid-word: flag those.
  • Rank by engagement relative to the account's own median, not absolute numbers. For the top and bottom posts (5 each with 15+ posts, 3 each below that), note the likely why: hook, format, topic, person featured, length, time.
  • Note efficacy or other claims in past captions that are not signed off, so drafts don't copy them.
  • Cadence actually kept: posts per week over the window you have (up to ~8 weeks). Say how long the window is.
  • Best posting slots: day and hour (in the account's time zone; ask it if unknown, since the server gives UTC) of the top performers. With fewer than ~10 posts, use the account's own usual posting times as the starting guess (label it so), not generic charts.
  • Competitor and category references: gooseworks:social_inspiration_search, then brand_read organic.competitor_top_posts. These are for formats and hooks only; never copy them. Skip any that are clearly outside the brand's category (a mis-resolved competitor).

3. Profile hygiene (text the human can paste)

  • Bio and pinned posts: no tool returns these today, so read the public profile in the host browser; if it won't open, ask in the draft-report message. Then: does the bio say who it's for and why to follow, in the voice? If not, draft one rewrite within the platform limit (IG 150, TikTok 80, X 160, LinkedIn headline 220).
  • Link in bio / pinned posts: do they point at the current priority? Suggest the swap.

Output

Account audit — <brand> · <date>
| Account | Platform | Job | Cadence kept (8 wks) | Median engagement | Best slots | Voice card? |
Top 5 / bottom 5 per account, with the why
What to make more of / stop: …
Bio + pinned suggestions: …
Unknowns asked: …

Write it to social/audit/<YYYY-MM-DD>.md. Write the account roster (jobs, handles, time zone, posting tool, free windows) to social/accounts.md. Both use gooseworks:file_write in the brand coworker scope.