agentleFS
moveo-aimoveo-ai/agent-skills/skills/build-agent/SKILL.md

build-agent

Skill1 starsChanged 50 days agoMITView on GitHub

Builds or changes a Moveo.AI conversational AI agent. Use when the user wants to create an agent from a brief or to write or fix its guidelines, tone or escalation rules. Also use to add intents and workflows, or to train, publish or roll back an agent. Requests such as "make the bot handle refunds" or "hand refunds over to a human" on a Moveo agent also qualify. Also matches Portuguese or Greek requests, such as "criar um agente" or "φτιάξε έναν agent".

9 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

11 rules. 0 are common, 2 are distinctive, and 9 are very specific to this project.

  • 0

    Common

    Most of the 9 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 9

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    If moveo_whoami did not confirm the account in this conversation, run the get-started skill first.

    line 8
  2. Very specific:

    Read moveo://schemas/guideline-authoring before you write any guideline text.

    line 19
  3. Very specific:

    Read moveo://schemas/agent-languages and moveo://schemas/agent-types.

    line 20
  4. Very specific:

    Call moveo_create_agent with the name, language, agent type and the guidelines distilled from the brief, in one call.

    line 21
  5. Very specific:

    Call moveo_validate_agent_guidelines and fix what it reports.

    line 23
  6. Very specific:

    Test the draft with the test-agent skill before you publish.

    line 25
  7. Very specific:

    Read before you write: moveo_get_agent_guidelines for the playbook, moveo_list_agent_workflows for the flows.

    line 33
  8. Very specific:

    Call moveo_publish_agent once, and wait for it to return.

    line 42
  9. Very specific:

    If the new version ends in status: 'failed', fix the content, wait for training, and publish again.

    line 43
  10. Distinctive:

    If guidelines repeat the knowledge base, answers go stale after documents change.

    line 53
  11. Distinctive:

    Before each publish, rollback or delete, name the account, the resource and what will change, then wait for an explicit yes.

    line 59

“Common” is measured: kept by at least 6 in 10 of the 9 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Build a Moveo.AI agent

If moveo_whoami did not confirm the account in this conversation, run the get-started skill first.

The model

  • An agent has one editable draft, version 0, and a list of immutable published versions. Only the draft accepts edits. Customers get the latest published version.
  • Guidelines are the natural-language playbook: goal, overview, features, objections, tone, handover rules. The guideline text becomes the agent's runtime prompt.
  • A workflow is a dialog plus the intents its nodes reference. Use workflows for flows that must happen exactly the same way every time, such as collecting an order number before a refund. Use guidelines for everything else.
  • Training starts on its own after an edit to intents, dialogs, guidelines or tools.

Build a new agent

  1. Read moveo://schemas/guideline-authoring before you write any guideline text. It gives the purpose and character budget of each field.
  2. Read moveo://schemas/agent-languages and moveo://schemas/agent-types. If the brief does not state the language, ask the user. The language field is permanent.
  3. Call moveo_create_agent with the name, language, agent type and the guidelines distilled from the brief, in one call.
  4. For each flow that must be exact, read moveo://schemas/workflow-node-types and moveo://guides/workflow-authoring, then call moveo_create_agent_workflow. See references/workflows.md.
  5. Call moveo_validate_agent_guidelines and fix what it reports.
  6. Poll moveo_get_agent until status is available.
  7. Test the draft with the test-agent skill before you publish.
  8. Publish only after the user agrees. See "Publish" below.

The build_agent_from_brief server prompt runs steps 1 to 8 in one pass. If the user has a complete brief and wants no questions, offer it.

Change an existing agent

  1. Find the agent in the agents snapshot (see get-started for its uri). If the name is ambiguous, ask.
  2. Read before you write: moveo_get_agent_guidelines for the playbook, moveo_list_agent_workflows for the flows.
  3. If the change touches more than one workflow, moves a jump, or changes a handover or a tag, show a plan before the first write. Give the current state, each change as before and after, and the nodes in other workflows that point at what you change (moveo_get_dialog_references). Make no change until the user agrees to the plan.
  4. Make the smallest edit that does the job. moveo_update_agent_guidelines deep-merges, so send only the fields you change.
  5. Show the user what changed with moveo_diff_agent_versions from the latest published version to 0.

Publish

  1. Poll moveo_get_agent until status is available.
  2. Show the diff from the latest published version to 0 and get the user's approval.
  3. Call moveo_publish_agent once, and wait for it to return. Do not call it again while it runs.
  4. If the new version ends in status: 'failed', fix the content, wait for training, and publish again. If the user wants the previous behavior back, offer moveo_rollback_agent_to_version.

Failure modes

  • Publish during training serves a half-trained agent. Always wait for status: 'available'.
  • moveo_update_agent edits top-level fields such as name, model, knowledge base and tools. Guideline fields sent to it are wrong. Use moveo_update_agent_guidelines for the playbook.
  • An edit to a published version fails. Edit the draft, then publish.
  • Rollback overwrites the current draft. Diff first. If the draft holds unpublished work, warn the user.
  • moveo_clone_agent works only on an agent that was never published.
  • Deleting an intent leaves dialog nodes that point at a slug that no longer exists. Call moveo_get_dialog_references first, then name what will break and wait for the user's yes.
  • If guidelines repeat the knowledge base, answers go stale after documents change. Put facts in the knowledge base and behavior in the guidelines.
  • An agent answers from a knowledge base only with two settings: the knowledge base is attached, and the search_knowledge_base tool is enabled. Use the knowledge-base skill for that.

Rules for every Moveo task

  • Reply in the user's language. Keep tool arguments, resource names and the product terms environment, agent and knowledge base as the tools expect them.
  • Before each publish, rollback or delete, name the account, the resource and what will change, then wait for an explicit yes. This holds even when the user asked for the change in the same message, because a request can name the wrong account or resource.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.