Queue work and run AI actions
Save writing, research, review or planning work for an external agent, run a chosen queue one action at a time, and review durable results.
On this page
When to use it
Use this skill to add an action for later, run the next action or a selected batch, answer blocked actions, and review completed reports. The ownpost entry skill includes this workflow too.
AI actions always run through your connected external client. They need no OpenAI or Anthropic provider key and do not use the app's saved provider credentials.
What you need
The current OwnPost app with the AI actions migration, the current OwnPost plugin (or updated skills), and an MCP connection with Full workspace tools. The legacy preparation-only scope cannot access general actions.
Install or select the skill
The ownpost entry skill includes this workflow. Use the command above if you want ownpost-actions as a separate shortcut, or select it from your installed plugin.
Try it
Select the skill in your connected assistant, then adapt these examples to your own content.
How it works
Name the workspace account and whether you want to add work for later or run existing work. An add-only request saves a Queued action and stops. Choose saved drafts only when draft creation is part of your request; research, reviews and plans normally return reports.
Use contextual requests for Idea discovery, selected-Post reviews, Calendar plans, Insights and writing improvements. The saved brief retains the selected IDs, topics, filters and dates; actions can also link selected blogs for research or review.
For Next action, the assistant claims at most one item. For selected or all-currently-queued work, it freezes the ordered IDs, revisions and one account, then completes one action before claiming another. Later additions are excluded.
The agent reads the pinned brief and approved context and uses the tools available in its client. Missing research capabilities or unclear instructions must be reported honestly. A Needs input question, failure, interruption or conflict stops the batch.
During longer work, the assistant can save partial report progress and next steps. Blog findings remain report proposals; applying article edits uses a separately requested blog workflow.
Open the saved action result in Assistant → AI actions. Review its report, sources, proposed text and actual saved draft links. Completion saves requested drafts with the result; proposed rewrites remain suggestions until you choose to apply them.
Answer a Needs input question to requeue work, then request a run when ready. Review failed or interrupted work before explicitly retrying. Recover an uncertain result before repeating work, and use Stop and edit when changing a running action.
Saving an action does not launch a client, schedule an agent or spend provider API credits. Completing an action does not publish to X or accept a writing-style change. Personal routines and legacy preparation jobs are separate.
Check the result
A saved queue item, completed report with output links, or a clear question/error appears in the selected account. A next-action request ends after one item.
If something does not work
If action tools are missing, check Full workspace tools access, update both the app and plugin or skills, then refresh tool discovery or reopen the client. A copied request must be pasted into the connected client. A saved connection or last-use time does not prove an agent is online.