Skip to content

SkedCast for AI agents

AI marketing agent prompts for automated social media posting

An AI marketing agent handles automated social media posting through real example prompts you can copy for Claude or any MCP client, built against SkedCast’s 59 documented MCP tools — draft, check, and schedule without writing a line of code.

How does an AI marketing agent handle automated social media posting?

Each prompt below maps to real, documented MCP tools — the same catalog listed on the AI agents overview — so the agent plans, drafts, checks, and schedules posts using the exact scope your account already has, never more.

Prompts by what you’re trying to do

Plan and publish

Draft, check, and schedule a post — or save it as a draft — across the accounts you choose.

"Draft a launch announcement for our Instagram and LinkedIn accounts, check it against both platforms’ rules, and schedule it for Tuesday at 9am if it passes."

  • validate_post
  • create_post

"Move the post I scheduled for Thursday on the Acme LinkedIn Page to Friday at the same time."

  • list_scheduled_posts
  • reschedule_post

"Cancel the TikTok post that was about to go out for the Nova account — we’re pulling it."

  • list_post_targets
  • cancel_post

Run the calendar

Answer "what goes out this week", move a whole campaign, or retry what failed.

"What’s scheduled for the Acme client this week, across every platform?"

  • list_post_targets

"Shift every pending post in the product-launch campaign back by two days."

  • list_campaigns
  • shift_campaign

"Show me every post target that failed in the last 24 hours and retry the ones that failed on a rate limit."

  • list_post_targets
  • retry_post

Pick the right time

Read each account’s posting limits, best times and composing rules before choosing a slot.

"What are the best times to post for our Instagram account, and how close is it to its daily posting limit?"

  • get_best_times
  • get_posting_limits

"Before I draft a TikTok video post, what are the caption, hashtag and video-length rules for TikTok?"

  • get_capabilities

"How much of this week’s posting quota do we have left before we hit the scarce-bucket limit?"

  • get_quota

Learn from results

Pull workspace, client, account and post analytics, and reports your team already set up.

"Give me last month’s reach, impressions and engagement, broken down by platform."

  • get_analytics_overview

"How did the three posts in the product-launch campaign perform, post by post?"

  • get_campaign
  • get_post_analytics

"Pull up the client report we already have set up for Acme and tell me if the latest run finished."

  • list_reports
  • get_report

Work with media

Find assets already in the library or bring new ones in from a link.

"Find every publishable video in the library tagged 'product-launch' and list their durations."

  • list_library_assets

"Import this image from our CDN into the library so I can attach it to a post: https://cdn.example.com/launch-hero.jpg"

  • import_media

Keep humans in charge

See the approval queue and record a decision — only when the connected person is allowed to.

"What’s waiting in the approval queue for the Acme client right now?"

  • list_approvals

"Show me the pending approval for the Instagram carousel, then approve it if the caption looks right."

  • get_approval
  • decide_approval

See how the connection works and what’s safe to hand over

Agent prompts FAQ

Do these prompts work with any MCP client, or only Claude?
Any client that speaks the Model Context Protocol over Streamable HTTP with OAuth can use these prompts — Claude included. You add https://api.skedcast.com/mcp as a remote MCP server and sign in once.
Will the agent post something without confirming first?
No — 28 of SkedCast’s 59 MCP tools change something, and every one is marked as a write tool so a well-behaved host confirms with you before calling it, the same as scheduling a post from create_post or approving one with decide_approval.
Can I write my own prompt instead of copying one of these?
Yes — these are starting points, not a fixed script. Describe what you want in plain language and the agent picks the matching tool from the same catalog these examples were built from.
Do these tool names match what the MCP server actually registers?
Yes — every tool name here is taken directly from the MCP tool catalog the server publishes, the same one rendered on the developer docs and in llms.txt, so nothing here is invented.
What can the agent NOT do through these prompts?
Only what the connected person’s own role and plan already allow — an agent acts as you, with your client access and your workspace’s approval rules, never with more reach than your own account.
Where do I see the full list of tools, not just these examples?
The developer docs and the /ai page both list the complete tool catalog with the scope each one needs.

Connect an assistant and try one of these

Add SkedCast as a remote MCP server, sign in, and ask for what you need.