playbooks

home / skills / nweii / agent-stuff / extract-flow-scenario

extract-flow-scenario skill

/skills/extract-flow-scenario

This skill extracts a precise workflow narrative from conversation into a structured markdown flow for accurate design and specification.

npx playbooks add skill nweii/agent-stuff --skill extract-flow-scenario

Review the files below or copy the command above to add this skill to your agents.

Files (1)
SKILL.md
2.2 KB
---
name: extract-flow-scenario
description: Extracts a sequence of workflow events from conversation context into a structured markdown list. Surfaces actors, state changes, and pain points to serve as raw data for any later synthesis, design planning, or spec development. Use when asked to document an actual operational flow, user journey, or scenario from a conversation to capture exactly what happened.
argument-hint: [flow-scenario]
disable-model-invocation: true
metadata:
  author: nweii
  version: "1.0.0"
---

When asked to document a workflow or scenario, output a fenced markdown code block containing a numbered list. Rules:

- **Title:** Begin with a concise descriptive heading capturing the subject, action, and key conditions (e.g., "[Entity/Subject] [Specific Action] ([Key Context/Condition])").
- **Specificity:** Document the _actual, specific_ scenario that occurred. Do NOT generalize or abstract into a generic workflow. This is a single building block. Include real names, identifiers, tool/platform names, and exact values where known.
- **Inline Structure:** Make retrieval easier by using bold formatting or inline tags (e.g., `Trigger:`, `Actor:`, `Outcome:`, `State Change:`) to call out when the flow starts, who does what, and when a state handoff occurs.
- **Optional Metadata:** If the user explicitly asks for a metadata block (or "properties"), include a brief block at the top of the flow detailing the `Trigger`, `Actors`, and `Outcome`. Ignore this block unless requested.
- **Structure:** Use nested indents (4 spaces) for sub-steps, alternative paths, and conditional branches (e.g., "If X... / If Y..."). Keep step language terse and direct.
- **Epistemological Honesty:** Document _only_ what was described. If you notice a logical gap or missing technical step in the context, do not silently invent it. Instead, **pause and ask the user for clarification in the conversation** before generating the final markdown block. This saves tokens, makes the process more collaborative, and prevents the need to continuously regenerate complex outputs.
- **Pending States:** Leave open-ended flows with a final pending step noting exactly what input or event is awaited before the flow can continue.

Overview

This skill extracts a concrete sequence of workflow events from a conversation and emits a ready-to-use, structured markdown numbered list in a fenced code block. It surfaces actors, explicit state changes, triggers, and pain points exactly as described — no abstraction or invented steps. Use it to capture a single observed scenario as raw input for downstream synthesis, design, or spec work.

How this skill works

When given a conversational transcript or scenario description, the skill parses the text for specific, named actors, tools, timestamps, and explicit actions. It builds a terse numbered list with inline tags (Trigger:, Actor:, State Change:, Outcome:) and uses nested indents for sub-steps and conditional branches. The skill refuses to invent missing technical steps; instead it pauses and requests clarification if a logical gap is detected. The final output is a fenced markdown code block containing the precise, documented flow.

When to use it

  • You need an exact record of an operational incident or user journey from a conversation.
  • Preparing raw inputs for UX, engineering specs, or incident postmortems.
  • Converting meeting notes or chat logs into a single, verifiable flow artifact.
  • Capturing flows that include exact tool names, IDs, or values for reproducibility.
  • Documenting a scenario where you must preserve what was actually said, not a generalized process.

Best practices

  • Provide the full conversation context or transcript so the skill can extract exact names, values, and timestamps.
  • State explicitly if you want an optional metadata/properties block (Trigger, Actors, Outcome).
  • Point out any suspected gaps in the conversation before extraction so the skill can ask for clarification rather than guessing.
  • Keep requested scope to one scenario per extraction to preserve specificity and traceability.
  • Review the generated flow for missing facts and respond with clarifying details if the skill indicates a pause.

Example use cases

  • Convert a Slack incident thread about a failed deployment into a step-by-step flow with actors and state changes.
  • Turn a customer support chat into a reproducible user-journey artifact noting pain points and where handoffs occurred.
  • Document a demo run where specific feature flags, environment names, or IDs were used.
  • Capture a multi-step approval process described in a meeting, preserving who approved and when.
  • Extract a troubleshooting conversation to isolate the exact symptom sequence and pending next action.

FAQ

Will the skill invent missing technical steps?

No. If the conversation omits a required step or contains a logical gap, the skill will pause and ask you for clarification before producing the final markdown.

Can it include an early metadata block?

Yes — include an explicit request for a metadata/properties block (Trigger, Actors, Outcome) and the skill will place it at the top of the flow.