🤖 AI Tools
LangGraph Node Sketch from a Workflow Brief
Sketch LangGraph nodes and edges from a workflow brief. No invented package versions or API keys.
0Reviews
Prompt
Act as a LangGraph workflow sketcher who only uses a pasted workflow brief. You sketch nodes, edges, and state fields for LangGraph from a workflow brief. You do not invent package versions, API keys, or tools not listed. This is not a full runnable app and not a LangChain agent dump unrelated to the brief. You work only from Inputs. Do not invent stats, citations, quotes, URLs, names, IDs, or records that are not in Inputs. Inputs: - Workflow brief: [Brief] - LangGraph version I lock (or UNKNOWN): [Version] - State fields I allow: [State] - Tools or retrievers I may name: [Tools] - Human-in-the-loop points: [HITL] - Words I must not use: [Banned] - What I must never invent: [Never] - Failure modes I care about: [Fails] - Output format: [Format] - Language for stubs (Python/TS): [Lang] Generate: 1. Honesty ledger: Brief, Version, State, Tools, HITL, Fails, Lang. Forbidden: keys, versions not locked, extra tools. 2. Node list with one-line purpose each from Brief. 3. Edge list including conditional edges only if Brief states a branch. 4. State schema using State fields only. 5. HITL breakpoints only as listed. 6. Stub signatures in Lang without real credential values. 7. Refuse: inventing model names not in Brief. 8. Compliance pass: cut Banned and Never. Format as Format. Constraints: - LangGraph sketch from Brief. Not a production deploy guide. - Never invent API keys or package versions when Version is UNKNOWN. - Only Tools named in Inputs. - No emojis.
Instructions
Replace every [bracket] with your details before running. Works on ChatGPT, Claude, and Gemini.
Examples
Example Input
Brief: intake email -> classify billing vs product -> retrieve FAQ -> draft reply -> human approve -> send Version: langgraph UNKNOWN (do not invent) State: raw_email, label, faq_hits, draft, approved Tools: faq_retriever; send_email (names only) HITL: approve draft before send Banned: sk- invent, model name invent if not in Brief Never: invent Redis checkpointer config Fails: empty FAQ, human reject Format: mermaid + state bullets + stub names Lang: Python
Example Output
1. Ledger. Flow: email->classify->retrieve->draft->HITL approve->send. Version UNKNOWN. State: raw_email, label, faq_hits, draft, approved. Tools: faq_retriever, send_email. HITL approve. Fails: empty FAQ, reject. Lang Python. Forbidden: sk-, model invent, Redis invent. 2. Nodes. intake_parse; classify_label; retrieve_faq; draft_reply; human_approve; send_mail. 3. Edges. intake->classify->retrieve->draft->human_approve; human_approve approved=true -> send; approved=false -> draft. Empty FAQ branch: retrieve->draft with faq_hits empty flag (Fails). 4. State. Only the five fields. No extras. 5. HITL. human_approve interrupt before send_mail. 6. Stubs. def classify_label(state): ...; def retrieve_faq(state): ...; credentials via env names only, no values. 7. Refuse. Model name NOT IN BRIEF. Checkpointer NOT IN INPUTS. 8. Compliance. Cut banned. Format mermaid+bullets+stubs. Gaps: Version, model, checkpointer. 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.