CrewAI Role and Task YAML from a Multi-Agent Brief

Prompt

Act as a CrewAI YAML sketcher who only uses a pasted multi-agent brief. You write role and task YAML for CrewAI from the brief. You do not invent package versions, API keys, tools not listed, or agents not named. This is not a LangGraph node sketch and not a runnable deploy.
You work only from Inputs. Do not invent stats, citations, quotes, URLs, names, IDs, or records that are not in Inputs.

Inputs:
- 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]

Generate:
1. Honesty ledger: Brief, Version, Agents, Tools, Process, HITL, Lang. Forbidden: keys, extra agents, extra tools, versions not locked.
2. Role YAML: one agent block per Agents entry. Role, goal, backstory from Brief only. No model SKU if not in Brief.
3. Task YAML: one task per job in Brief. description, expected_output, agent from Agents. No secret values.
4. Process: only Process. If UNKNOWN, write process: UNKNOWN and do not pick hierarchical by default.
5. Tools: attach only Tools named. If a task needs a tool not listed, write NOT IN INPUTS rather than adding one.
6. HITL: insert a human review note only at HITL points.
7. Refuse: inventing OpenAI model names, Redis memory URLs, or API keys.
8. Compliance pass: cut Banned and Never. Print that Version is quoted or UNKNOWN. Format as Format.

Constraints:
- 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.

Instructions

Replace every [bracket] with your details before running. Works on ChatGPT, Claude, and Gemini.

Examples

Example Input

Brief: three-agent FAQ crew: Researcher gathers pasted policy bullets; Writer drafts FAQ answers; Reviewer checks quotes. No web browse. Output is FAQ markdown. Human approves before publish.
Version: crewai UNKNOWN (do not invent)
Agents: researcher (gather bullets); writer (draft FAQ); reviewer (quote check)
Tools: none named (do not invent a search tool)
Process: sequential as pasted
Banned: sk-, gpt-4o invent if not in Brief, redis:// invent, langgraph
Never: invent API keys, invent a browse tool, invent a fourth agent
Format: yaml blocks + refuse + gaps
Lang: English
HITL: reviewer output before publish

Example Output

1. Ledger. FAQ crew: researcher, writer, reviewer. Version UNKNOWN. Tools none named. Process sequential. HITL: reviewer before publish. Lang English. Forbidden: sk-, invented gpt-4o, redis://, langgraph, browse tool, fourth agent.

2. Role YAML.
agents:
- name: researcher
role: Policy-bullet gatherer
goal: Gather pasted policy bullets only
backstory: Uses pasted policy text. No web browse. Model SKU NOT IN BRIEF.
- name: writer
role: FAQ drafter
goal: Draft FAQ answers from gathered bullets
backstory: Quotes bullets. Does not invent SLAs.
- name: reviewer
role: Quote checker
goal: Check that answers quote pasted bullets
backstory: Flags unquoted claims. Human review next.

3. Task YAML.
tasks:
- name: gather_bullets
description: Collect pasted policy bullets. No browse.
expected_output: Bullet list quoted from paste
agent: researcher
- name: draft_faq
description: Draft FAQ markdown from bullets
expected_output: FAQ markdown with quotes
agent: writer
- name: check_quotes
description: Check quotes; flag unquoted claims
expected_output: FAQ plus flags
agent: reviewer

4. Process. sequential (as pasted). Not hierarchical. Not UNKNOWN.

5. Tools. none named. Do not attach search or browse. If a task wants search: NOT IN INPUTS.

6. HITL. After reviewer task, human approve before publish. No other interrupts in Brief.

7. Refuse. API keys NOT IN INPUTS. Model name NOT IN BRIEF. Redis URL NOT IN INPUTS. Fourth agent refused. LangGraph refused (wrong surface).

8. Compliance. Version UNKNOWN quoted. Banned cut. Format yaml+refuse+gaps. Gaps: CrewAI version, model SKU, tool list, memory config, publish channel.

Missing-data policy: if a field was blank, write NOT IN INPUTS rather than guessing. Lock any tool version named in Inputs; if unnamed, write unknown. No invented testimonials, star ratings, or press logos. If legal, clinical, insurance, HR, education-plan, or veterinary content appears, add a one-line not-advice and de-identify banner. Quote banned-word hits and cut them. End with a gaps list of five bullets the user still owes you. Character and byte caps in the job are hard; print counts when relevant. Refuse to backfill DOIs, exam dumps, PHI, PII, or compensation promises not in Inputs.