Antwork audit
Agent Skills for the Antwork MCP connector
npx -y skills add iker-gonzalez/antwork-skills --skill antwork-auditAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Use when the user wants a full audit, health check, or scorecard of their social-media presence managed through Antwork — engagement performance, voice consistency, posting cadence, content quality, and platform coverage. Trigger on "audit my socials", "how's my LinkedIn/X doing", "social health check", "review my Antwork account", "what should I fix", "grade my content".
SKILL.md
4.2 KB, as published. Nobody here has run it
Antwork Social Presence Audit
A comprehensive, data-driven audit of everything the user ships through Antwork. It mirrors a 5-dimension marketing audit: five specialized agents run in parallel, each owning one dimension, then you synthesize a single scored report.
1. Establish scope first
Before spawning anything, gather the lay of the land so the agents (and the report) reference real accounts and real numbers:
whoami— confirm the user and default workspace.list_workspaces— if more than one and no default, ask which to audit (orset_default_workspace).list_social_accounts— the connected platforms, handles, and token health. This is the audit's universe.get_workspace_settings— timezone, preferred posting times, brand identity.
If no accounts are connected, stop and route the user to antwork-setup instead — there's nothing to audit yet.
2. Spawn the five dimension agents in parallel
Launch all five in a single batch so they run concurrently. Each is read-only and returns a structured findings block with a 0–100 sub-score and its top 3 fixes. Pass each the workspace and the account list you gathered.
| Dimension | Agent | Weight | Primary MCP tools |
|---|---|---|---|
| Performance | antwork-performance | 30% | get_performance, get_engagement_history, get_optimal_posting_times |
| Voice Consistency | antwork-voice-analyst | 20% | get_post_context (per account), list_posts |
| Cadence & Timing | antwork-cadence | 20% | list_posts, get_calendar, get_optimal_posting_times |
| Content Quality | antwork-content | 20% | fetch_platform_posts, get_post, list_posts |
| Platform Coverage & Growth | antwork-growth | 10% | list_social_accounts, get_workspace_settings |
Do not analyze the dimensions inline yourself — the parallel agents are the point. If one agent fails or returns nothing, note the gap and score that dimension conservatively rather than blocking the report.
3. Compute the composite Social Health Score
Weighted sum of the five sub-scores (0–100):
Score = Performance·0.30 + Voice·0.20 + Cadence·0.20 + Content·0.20 + Growth·0.10
Band it: 80–100 Strong · 60–79 Healthy · 40–59 Needs work · 0–39 At risk.
4. Write ANTWORK-AUDIT.md
Produce a client-ready Markdown file with:
- Header — workspace name, audited accounts (platform + handle), date, and the composite score with its band.
- Executive summary — 3–5 sentences: the single biggest strength, the single biggest leak, and the one move with the highest payoff.
- Scorecard table — each dimension, its sub-score, and a one-line verdict.
- Per-dimension findings — one section per agent, carrying through its findings and top-3 fixes verbatim-in-spirit (don't dilute the specifics).
- Prioritized action plan — a single merged, impact-ranked checklist drawn from all five agents' fixes. Each item: what to do, which Antwork skill/tool does it (
antwork-voiceto refresh a stale profile,antwork-calendarto fill cadence gaps, etc.), and expected payoff.
5. Close with the next move
End by offering to execute the top action immediately through the relevant skill — e.g. "Want me to refresh the stale X voice profile now?" (antwork-voice) or "Want me to fill next week's empty slots?" (antwork-calendar). The audit is only valuable if it turns into shipped posts.
Guardrails
- The audit is read-only. Never create, edit, schedule, publish, or delete anything during an audit.
refresh_post_metricsburns platform API quota — only suggest it; don't have agents call it during a routine audit unless the user explicitly wants live-refreshed numbers.- Quote real numbers from the analytics rows. Never fabricate engagement figures or invent posts that aren't in the data.