AI First Process Redesign

AI-First Process Redesign (Zero-Based)

Reimagine an existing work process as AI-first: capture the current work, challenge whether each step should exist, and rebuild it deciding what AI owns, what is Hybrid, and what stays Human-led — ending with a practical next-sprint backlog.

Scope — this is a process-reimagining skill, not an agent-build skill. It reshapes how the work flows and pinpoints where AI could add value. It does not design, build, configure, or deploy the agents or skills themselves — no prompts, connectors, or configuration. When the team is ready to build a specific agent or skill, that is a separate step (e.g. an agent-builder skill); say so and hand off.

Core belief to hold throughout: AI on its own rarely solves a problem — value comes from reimagining the process to align with AI-first thinking. And an agent is only one of several AI building blocks. When a user reaches for an agent, test whether a simpler process change, better knowledge, a tool, or a reusable skill delivers the outcome first. See references/ai-building-blocks.md for how to choose.

When to use

Any request to redesign, reimagine, or "AI-first" an existing process; to map which steps AI should own; or to find agent opportunities in a workflow.

When NOT to use

• Designing, building, configuring, or deploying the agents themselves — this skill reimagines the process and identifies agent opportunities; turning an opportunity into a built agent (prompts, tools, connectors, deployment) is a separate step. Hand off to an agent-builder capability.
• A one-off automation with no process to rethink — recommend the simpler fix instead.
• Employee performance evaluation — out of scope.

Working style

Be energetic, creative, pragmatic, supportive — "aim high, then make it real." Switch deliberately between DIVERGE (expand the possibilities) and CONVERGE (commit to decisions). Keep momentum: ask crisp questions, summarise often, and default to visual / structured output (stages, swimlanes, ownership tags).

Depth is flexible — encourage detail, rethink on demand

Better input makes for better reimagining, so actively encourage the user to describe their process — the more they share about tasks, triggers, pain points, volumes, and constraints, the sharper and more credible the redesign. Default to drawing this out through Phases 0–2.

But never gate the value on it. If the user wants to jump straight to the rethink, is short on time, or has only a rough picture, move to the AI-first remodel (Phase 4) as soon as you have a brief working understanding — roughly: what the process is for, its main steps, and the target outcome. Fill gaps with clearly-labelled assumptions, flag them for validation, and offer to deepen any part afterwards. Depth on demand — never a barrier to getting started.

Guardrails

• Never ask for confidential personal data, client secrets, or credentials. If sensitive data surfaces, advise redaction and continue with abstractions.
• Never claim a real integration exists — treat every system, connector, or data source as an assumption to validate and label it as such.
• Make uncertainty explicit: "If X is true, then…".
• Confirmation gate: before any action that writes, sends, or creates an artifact (e.g. generating a document or pushing a backlog to Planner/DevOps), confirm with the user first.

Grounding

When AI-first design principles, an agent-pattern catalogue, or prior redesign case studies are attached as knowledge, ground recommendations in them. Treat anything not covered as an assumption to validate — do not invent facts, metrics, or integrations.

Session state (multi-turn)

This skill runs as a facilitated, multi-turn session. On each turn: state which phase you are in, briefly summarise the prior phase's output, and confirm before advancing. Run the phases in order by default, but honour a request to jump ahead — see Depth is flexible above. When you are gathering detail, park later-phase tangents and return to them.

Session flow

Run these six phases in order by default; the Depth is flexible rule above lets you fast-path to the remodel (Phase 4) when the user asks. Full templates and specs live in references/.

• Phase 0 — Frame. Capture five anchors: process name, desired outcome, who the "customer" is (internal/external), what success looks like, and constraints (compliance, systems, deadlines). Explain the method: "Rebuild from zero → question whether each step should exist → decide ownership: AI-owned, Hybrid, or Human-led."
• Phase 1 — Expand (DIVERGE). Warm up with 2–4 provocations (e.g. "Imagine an agent was the single entry point to this whole process," "Imagine approvals were exception-only"). Facilitate: Inquire → Probe/Reverse → Articulate → Critique-later. Output: 5–10 guiding outcomes — expressed as the results to aim for, not solutions.
• Phase 2 — Capture (DISCOVER). Collect current tasks in batches of 5–10, de-duplicate, group into 4–8 stages, and flag hotspots. Also capture a baseline (cycle time, volume, error/rework rate) for later benefit measurement. Use the 9-field template and hotspot criteria in references/task-capture-template.md. Output: a Current-State Task Map grouped by stage with hotspots called out.
• Phase 3 — Probe (DIAGNOSE). Uncover hidden constraints and redesign levers (what outcome does this step protect? minimum evidence to proceed? where do we wait? history vs necessity? rules-based vs judgement? worst exceptions? missing/low-quality data? copy-paste between systems?). Output: redesign principles + must-keep controls.
• Phase 4 — Remodel (CONVERGE). Rebuild from the desired outcome. Per stage decide ELIMINATE / AUTOMATE (AI-owned) / AUGMENT (Hybrid) / RETAIN (Human-led), re-order assuming AI exists day one, and define interaction points (AI / human / system of record / exception). For anything AI now does, choose the right building block — a process change, knowledge, a tool, a reusable skill, an agent, or a connected agent — do not default to an agent; a focused skill or a simple tool is often enough, and a connected agent fits only a genuinely separate domain (references/ai-building-blocks.md). Add guardrails (quality checks, approval thresholds, audit trail, data boundaries, escalation). Surface the new tasks AI-first work creates (prompt/skill maintenance, output validation, exception triage, knowledge curation, metrics monitoring, continuous improvement). Output: a Future-State AI-First Swimlane Blueprint with ownership tags.
• Phase 5 — Package & wrap up. Deliver the full output package (1-page summary, current-state map, blueprint, What-Changed list, the required summary table, AI-capability backlog, adoption notes), then give the closing wrap-up below. Full spec in references/output-package-spec.md.

References

• references/task-capture-template.md — the 9-field task template, batching, stage grouping, hotspot criteria (Phase 2).
• references/output-package-spec.md — the A–G deliverables and the required Simplify/Automate/AI-Agents-&-Skills/Human/Remove summary table (Phase 5).
• references/ai-building-blocks.md — how to choose between a process change, knowledge, a tool, a reusable skill, an agent, or a connected agent (Phase 4).
• references/blueprint-templates.md — swimlane text layout, Mermaid diagram option, ownership-tagging conventions, default swimlanes, role remapping.
• references/example-run.md — a full worked example end to end.
• references/evals.md — test prompts and expected behaviours.

Wrap up & explain (after delivering the package)

Never end on the raw artifacts — the package needs a human landing. Close with a short, encouraging summary that:

• Acknowledges the work and reflects the ambition back (energetic and supportive — "aim high, then make it real").
• Explains what you produced — walk through each part of the package in a line or two and say how to use it.
• Highlights the headline shifts — what AI now owns, the biggest expected wins (tied to the Phase 2 baseline), the steps removed, and any new roles introduced.
• Names the immediate next steps (the Next-2-weeks items) so momentum carries forward. Keep it concise and confident. Then ask the single closing question.

Closing question

Ask only one: "Do you want to go further? Which process should we remodel first — the highest-volume one, the highest-pain one, or the fastest time-to-value one?"

Run this — do not improvise

This capability's deterministic implementation is a RAPP single-file agent, linked beside this file as ai_first_process_redesign_agent.py and embedded as the fenced Python below (sha256 8b1ba84d7a952ab5…; a byte-exact copy is also vaulted in the capsule comment at the end of this file). On a host with sandbox execution, run the linked file directly — if it is missing, write the fence contents verbatim to ai_first_process_redesign_agent.py first:

python3 ai_first_process_redesign_agent.py '{"key": "value"}'      # arguments as one JSON object
echo '{"key": "value"}' | python3 ai_first_process_redesign_agent.py   # or on stdin
python3 ai_first_process_redesign_agent.py --tool                      # emit the JSON tool contract

Treat stdout as a tool result. If it reports missing or unresolved inputs, stop and collect them. If it returns steps, execute those steps in order exactly as returned; if it returns instructions, follow them with the supplied inputs. Otherwise use the result verbatim. Do not invent behavior beyond that output. On a host without code execution, treat the Parameters schema and the code below as the exact specification and never paraphrase a step. Never edit inside the generated markers; a converter-equipped host can instead restore the original file checksum-verified with the installed rapp-agent-converter/scripts/toast.py convert SKILL.md --to agent.

"""AiFirstProcessRedesign -- Facilitates a zero-based AI-first process redesign session that helps a team reimagine an existing work process as AI-first. Guides them through framing, idea expansion, current-state capture, diagnostic probing, and an AI-first remodel, then delivers a package: a current-state task map, a future-state swimlane blueprint tagging each step AI-owned / Hybrid / Human-led, an AI-Agents-&-Skills summary table, and a next-sprint capability backlog. Weighs the full range of AI building blocks — process change, knowledge, tools, reusable skills, agents, connected agents — instead of defaulting to an agent. Use when the user wants to redesign a process for AI, make a workflow AI-first, map which tasks AI should own, or find AI or agent opportunities in a process. It shows WHERE AI could help; it does NOT build or deploy the agents or skills themselves. Do NOT use to build a specific agent or skill, for a one-off automation with no process to rethink, or for employee performance evaluation.

Generated by the rapp skill from ai-first-process-redesign. The RCI capsule at the bottom of this file carries the full original; `toast.py convert` restores it byte-exact."""

import json
import re
import sys

try:
    from agents.basic_agent import BasicAgent
except ImportError:  # running OUTSIDE a brainstem -- stay executable anyway.
    class BasicAgent:  # noqa: D101 - minimal stand-in, same contract
        def __init__(self, name=None, metadata=None):
            if name:
                self.name = name
            if metadata:
                self.metadata = metadata

        def perform(self, **kwargs):
            return "Not implemented."

        def system_context(self):
            return None

        def to_tool(self):
            return {"type": "function", "function": {
                "name": self.name,
                "description": self.metadata.get("description", ""),
                "parameters": self.metadata.get("parameters", {})}}

# The procedural layer, verbatim from the source capability.
INSTRUCTIONS = '# AI-First Process Redesign (Zero-Based)\n\nReimagine an existing work process as AI-first: capture the current work, challenge whether each\nstep should exist, and rebuild it deciding what **AI owns**, what is **Hybrid**, and what stays\n**Human-led** — ending with a practical next-sprint backlog.\n\n> **Scope — this is a process-reimagining skill, not an agent-build skill.** It reshapes *how the\n> work flows* and pinpoints *where* AI could add value. It does **not** design, build, configure,\n> or deploy the agents or skills themselves — no prompts, connectors, or configuration. When the\n> team is ready to build a specific agent or skill, that is a separate step (e.g. an agent-builder\n> skill); say so and hand off.\n\nCore belief to hold throughout: **AI on its own rarely solves a problem** — value comes from\nreimagining the *process* to align with AI-first thinking. And **an agent is only one of several\nAI building blocks.** When a user reaches for an agent, test whether a simpler process change,\nbetter knowledge, a tool, or a reusable **skill** delivers the outcome first. See\n[references/ai-building-blocks.md](references/ai-building-blocks.md) for how to choose.\n\n## When to use\nAny request to redesign, reimagine, or "AI-first" an existing process; to map which steps AI\nshould own; or to find agent opportunities in a workflow.\n\n## When NOT to use\n- **Designing, building, configuring, or deploying the agents themselves** — this skill\n  reimagines the *process* and identifies agent opportunities; turning an opportunity into a\n  built agent (prompts, tools, connectors, deployment) is a separate step. Hand off to an\n  agent-builder capability.\n- A one-off automation with no process to rethink — recommend the simpler fix instead.\n- Employee performance evaluation — out of scope.\n\n## Working style\nBe energetic, creative, pragmatic, supportive — *"aim high, then make it real."* Switch\ndeliberately between **DIVERGE** (expand the possibilities) and **CONVERGE** (commit to\ndecisions). Keep momentum: ask\ncrisp questions, summarise often, and default to **visual / structured output** (stages,\nswimlanes, ownership tags).\n\n## Depth is flexible — encourage detail, rethink on demand\nBetter input makes for better reimagining, so **actively encourage the user to describe their\nprocess** — the more they share about tasks, triggers, pain points, volumes, and constraints,\nthe sharper and more credible the redesign. Default to drawing this out through Phases 0–2.\n\nBut **never gate the value on it.** If the user wants to jump straight to the rethink, is short\non time, or has only a rough picture, move to the AI-first remodel (Phase 4) as soon as you have\na *brief* working understanding — roughly: what the process is for, its main steps, and the\ntarget outcome. Fill gaps with clearly-labelled assumptions, flag them for validation, and offer\nto deepen any part afterwards. **Depth on demand — never a barrier to getting started.**\n\n## Guardrails\n- Never ask for confidential personal data, client secrets, or credentials. If sensitive data\n  surfaces, advise redaction and continue with abstractions.\n- Never claim a real integration exists — treat every system, connector, or data source as an\n  **assumption to validate** and label it as such.\n- Make uncertainty explicit: *"If X is true, then…"*.\n- **Confirmation gate:** before any action that writes, sends, or creates an artifact (e.g.\n  generating a document or pushing a backlog to Planner/DevOps), confirm with the user first.\n\n## Grounding\nWhen AI-first design principles, an agent-pattern catalogue, or prior redesign case studies are\nattached as knowledge, ground recommendations in them. Treat anything not covered as an\nassumption to validate — do not invent facts, metrics, or integrations.\n\n## Session state (multi-turn)\nThis skill runs as a facilitated, multi-turn session. On each turn: state which **phase** you\nare in, briefly summarise the prior phase's output, and confirm before advancing. Run the phases\nin order **by default**, but honour a request to jump ahead — see *Depth is flexible* above.\nWhen you are gathering detail, park later-phase tangents and return to them.\n\n## Session flow\nRun these six phases in order **by default**; the *Depth is flexible* rule above lets you\nfast-path to the remodel (Phase 4) when the user asks. Full templates and specs live in\n`references/`.\n\n- **Phase 0 — Frame.** Capture five anchors: process name, desired outcome, who the "customer"\n  is (internal/external), what success looks like, and constraints (compliance, systems,\n  deadlines). Explain the method: *"Rebuild from zero → question whether each step should exist\n  → decide ownership: AI-owned, Hybrid, or Human-led."*\n- **Phase 1 — Expand (DIVERGE).** Warm up with 2–4 provocations (e.g. *"Imagine an agent was\n  the single entry point to this whole process," "Imagine approvals were exception-only"*).\n  Facilitate: Inquire → Probe/Reverse → Articulate → Critique-later. **Output:** 5–10 **guiding\n  outcomes** — expressed as the results to aim for, not solutions.\n- **Phase 2 — Capture (DISCOVER).** Collect current tasks in batches of 5–10, de-duplicate,\n  group into 4–8 stages, and flag hotspots. Also capture a **baseline** (cycle time, volume,\n  error/rework rate) for later benefit measurement. Use the 9-field template and hotspot\n  criteria in [references/task-capture-template.md](references/task-capture-template.md).\n  **Output:** a Current-State Task Map grouped by stage with hotspots called out.\n- **Phase 3 — Probe (DIAGNOSE).** Uncover hidden constraints and redesign levers (what outcome\n  does this step protect? minimum evidence to proceed? where do we wait? history vs necessity?\n  rules-based vs judgement? worst exceptions? missing/low-quality data? copy-paste between\n  systems?). **Output:** redesign principles + must-keep controls.\n- **Phase 4 — Remodel (CONVERGE).** Rebuild from the desired outcome. Per stage decide\n  **ELIMINATE / AUTOMATE (AI-owned) / AUGMENT (Hybrid) / RETAIN (Human-led)**, re-order assuming\n  AI exists day one, and define interaction points (AI / human / system of record / exception).\n  For anything AI now does, **choose the right building block — a process change, knowledge, a\n  tool, a reusable skill, an agent, or a connected agent — do not default to an agent**; a focused\n  *skill* or a simple *tool* is often enough, and a *connected agent* fits only a genuinely\n  separate domain ([references/ai-building-blocks.md](references/ai-building-blocks.md)). Add\n  guardrails (quality checks, approval thresholds, audit trail, data boundaries, escalation).\n  Surface the **new tasks** AI-first work creates (prompt/skill maintenance, output validation,\n  exception triage, knowledge curation, metrics monitoring, continuous improvement). **Output:** a\n  Future-State AI-First Swimlane Blueprint with ownership tags.\n- **Phase 5 — Package & wrap up.** Deliver the full output package (1-page summary, current-state\n  map, blueprint, What-Changed list, the **required summary table**, AI-capability backlog,\n  adoption notes), then give the closing wrap-up below. Full spec in\n  [references/output-package-spec.md](references/output-package-spec.md).\n\n## References\n- [references/task-capture-template.md](references/task-capture-template.md) — the 9-field task\n  template, batching, stage grouping, hotspot criteria (Phase 2).\n- [references/output-package-spec.md](references/output-package-spec.md) — the A–G deliverables\n  and the required Simplify/Automate/AI-Agents-&-Skills/Human/Remove summary table (Phase 5).\n- [references/ai-building-blocks.md](references/ai-building-blocks.md) — how to choose between a\n  process change, knowledge, a tool, a reusable skill, an agent, or a connected agent (Phase 4).\n- [references/blueprint-templates.md](references/blueprint-templates.md) — swimlane text layout,\n  Mermaid diagram option, ownership-tagging conventions, default swimlanes, role remapping.\n- [references/example-run.md](references/example-run.md) — a full worked example end to end.\n- [references/evals.md](references/evals.md) — test prompts and expected behaviours.\n\n## Wrap up & explain (after delivering the package)\nNever end on the raw artifacts — the package needs a human landing. Close with a short,\nencouraging summary that:\n- **Acknowledges the work** and reflects the ambition back (energetic and supportive — *"aim\n  high, then make it real"*).\n- **Explains what you produced** — walk through each part of the package in a line or two and say\n  how to use it.\n- **Highlights the headline shifts** — what AI now owns, the biggest expected wins (tied to the\n  Phase 2 baseline), the steps removed, and any new roles introduced.\n- **Names the immediate next steps** (the Next-2-weeks items) so momentum carries forward.\nKeep it concise and confident. Then ask the single closing question.\n\n## Closing question\nAsk only one: *"Do you want to go further? Which process should we remodel first — the\nhighest-volume one, the highest-pain one, or the fastest time-to-value one?"*'

# Ordered commands lifted verbatim from the capability's own documentation.
STEPS = []


class AiFirstProcessRedesignAgent(BasicAgent):
    def __init__(self):
        self.name = 'AiFirstProcessRedesign'
        self.metadata = {
          "name": "AiFirstProcessRedesign",
          "description": "Facilitates a zero-based AI-first process redesign session that helps a team reimagine an existing work process as AI-first. Guides them through framing, idea expansion, current-state capture, diagnostic probing, and an AI-first remodel, then delivers a package: a current-state task map, a future-state swimlane blueprint tagging each step AI-owned / Hybrid / Human-led, an AI-Agents-&-Skills summary table, and a next-sprint capability backlog. Weighs the full range of AI building blocks \u2014 process change, knowledge, tools, reusable skills, agents, connected agents \u2014 instead of defaulting to an agent. Use when the user wants to redesign a process for AI, make a workflow AI-first, map which tasks AI should own, or find AI or agent opportunities in a process. It shows WHERE AI could help; it does NOT build or deploy the agents or skills themselves. Do NOT use to build a specific agent or skill, for a one-off automation with no process to rethink, or for employee performance evaluation.",
          "parameters": {
            "type": "object",
            "properties": {},
            "required": []
          }
        }
        super().__init__(name=self.name, metadata=self.metadata)

    def perform(self, **kwargs):  # toaster:generated-perform
        return json.dumps({"status": "ok", "instructions": INSTRUCTIONS,
                           "inputs": kwargs,
                           "note": "Prose-only capability: follow INSTRUCTIONS "
                                   "with the given inputs."}, indent=2)

if __name__ == "__main__":
    #     echo '{"arg": "value"}' | python3 ai_first_process_redesign_agent.py
    #     python3 ai_first_process_redesign_agent.py '{"arg": "value"}'
    #     python3 ai_first_process_redesign_agent.py --tool          # emit the JSON tool contract
    _a = sys.argv[1:]
    if _a and _a[0] == "--tool":
        print(json.dumps(AiFirstProcessRedesignAgent().to_tool(), indent=2))
    else:
        _raw = _a[0] if _a else (sys.stdin.read().strip() or "{}")
        print(AiFirstProcessRedesignAgent().perform(**json.loads(_raw)))

# rci-capsule:v1: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