How to Write CrewAI Role and Task YAML from a Multi-Agent Brief
CrewAI YAML invents a browse tool and a model SKU. Version was UNKNOWN. Tools was none named. Leave keys as env names.
The matching generator is the CrewAI Role and Task YAML from a Multi-Agent Brief prompt. Browse related cards in the PromptDig library (Browse more prompts). When a filled run survives, share the version you actually use (Share a prompt).
One agent block per Agents entry
Write CrewAI role and task YAML from a multi-agent brief. No invented package versions, tools, or API keys. Start by filling Inputs, not by asking the model to remember last week's run. If a field is blank, write NONE or NOT IN INPUTS and leave it blank through Generate. The card is built so the model cannot honestly invent a number, owner, URL, or command that you did not paste.
Paste these fields before you hit run:
Multi-agent brief: [Brief]
CrewAI version I lock (or UNKNOWN): [Version]
Agent names and goals I allow: [Agents]
Tools I may name: [Tools]
Process type I allow (sequential/hierarchical or UNKNOWN): [Process]
Words I must not use: [Banned]
What I must never invent: [Never]
Output format: [Format]
Language for backstory stubs: [Lang]
Human review points: [HITL]
That inventory is the honesty ledger. Anything that does not appear there is forbidden in the draft. If you catch yourself adding a nice-to-have after the run, you are no longer using the card. You are ghostwriting. Put the extra fact in Inputs and run again.
Tasks only from jobs in the brief
Generate is numbered on purpose. Do not skip a step because the first paragraph looked done. The early steps exist to stop later prose from smuggling claims.
Walk the Generate list in order:
- Honesty ledger: Brief, Version, Agents, Tools, Process, HITL, Lang. Forbidden: keys, extra agents, extra tools, versions not locked.
- Role YAML: one agent block per Agents entry. Role, goal, backstory from Brief only. No model SKU if not in Brief.
- Task YAML: one task per job in Brief. description, expected_output, agent from Agents. No secret values.
- Process: only Process. If UNKNOWN, write process: UNKNOWN and do not pick hierarchical by default.
- Tools: attach only Tools named. If a task needs a tool not listed, write NOT IN INPUTS rather than adding one.
- HITL: insert a human review note only at HITL points.
- Refuse: inventing OpenAI model names, Redis memory URLs, or API keys.
- Compliance pass: cut Banned and Never. Print that Version is quoted or UNKNOWN. Format as Format.
If a step asks for a version lock, quote the version from Inputs in the output. If a step asks for a refuse list, keep the refuse list in the published artifact, not in a sidebar you delete. Reviewers should see what the model was not allowed to do.
Process UNKNOWN stays unknown
Most failures are the same shape: a missing field gets a confident fill. A conversion rate appears. A Gradle task appears. A flash point appears. A caption appears on a job that asked for slide text only. Your review is to search the draft for numbers, names, and commands, then grep Inputs. No match means cut.
Honor the constraints as hard stops, not vibes:
- CrewAI role and task YAML from Brief. Not a LangGraph sketch and not a production deploy.
- Never invent API keys or package versions when Version is UNKNOWN.
- Only Tools and Agents named in Inputs.
- No emojis.
When the card says not legal advice, not certification, not an exam dump, or not a caption engine, that sentence belongs at the top of the output. Deleting it to look more finished is how you inherit risk.
No keys and no extra tools
Finish with the compliance pass the prompt already asks for. Quote the banned-word hits. Cut them. Print character counts when the job has a cap. Print word counts when the job has a budget. List gaps as gaps. Five missing facts are more useful than one smooth paragraph.
Tags on the card (crewai role task yaml, crewai multi-agent brief, no invented crewai versions) are a reminder of the job shape, not an invitation to wander into a neighboring cluster. If you need a different surface, open a different PromptDig card rather than stretching this one.
Fill the card, then run
Replace every bracket. Run on ChatGPT, Claude, or Gemini. Read the ledger first, then the artifact. If the model invents a commit, KPI, DOI, PEL, bid, or logo, discard the run. Tighten Inputs. Run again. Share the filled card that survived, not the first draft that sounded done.