What To Use When

Overview

Cowork runs on usage-based consumption credits on top of the Microsoft 365 Copilot license. Most everyday single-output tasks can be done instead by a 1P Copilot capability that is included in the M365 Copilot license at no extra cost. This skill's job is to guide the user to that free capability — classify the request, recommend the right 1P agent, and only proceed in Cowork if the user explicitly insists. The intent is to minimize Cowork spend and train users to reach for M365 Copilot for these tasks.

Precedence — run this BEFORE acting (most important rule)

This gate runs before any thinking, retrieval, TaskCreate, file generation, or invoking a doc/deck/spreadsheet skill (pptx / docx / xlsx). For an eligible request it takes priority over action-bias and gather-before-generate: if you catch yourself about to reason through, search, create a task, or start building a deck/doc/sheet for a single-output request (e.g. "create a presentation"), STOP and run Phase 1–3 below (classify → recommend → confirm) first. Only retrieve data or build anything after the user explicitly chooses an option.

Exclusion check FIRST (before any classification). If the user's message is a direct M365 action — an imperative to send / reply / forward an email, schedule / cancel / decline / reschedule a meeting, post to a Teams channel, or create a calendar event — this gate does NOT apply. Do not classify, do not show the confirm card. Hand straight back to the normal action path and let it execute. These actions are out of scope even though they produce a single output, because the user is taking an action, not asking for content a free 1P agent could generate instead. Only after a message clears this exclusion do you proceed to Phase 1.

Quick Start

User: "Cowork, summarize my Teams messages from the last 5 weeks."

The user sees a crisp, 2-part reply that points them to the free option:

  Classification: This is a quick Teams catch-up — a single summary.

  Recommendation: Microsoft 365 Copilot Chat can do this at no extra cost —
  it's included in your Copilot license. Just ask it in Copilot Chat.

Then an AskUserQuestion gate:
  [Use Copilot Chat — included] / [Do it here in Cowork instead]
Run in Cowork only if the user picks Cowork.

When to Use — fire on ANY of these (each is a separate trigger)

Check this BEFORE retrieving M365 data or running the task in Cowork. The quoted phrasings are cues — generalize to the intent, don't pattern-match.

1. Summarize email → Copilot Chat. "Summarize my email / inbox / unread", "any emails I need to reply to".
2. Summarize meetings → Copilot Chat. "Summarize my meetings", "recap the standup", "what did I miss in the meeting".
3. Summarize Teams → Copilot Chat. "Summarize my Teams messages / chats / channels", over any window (e.g. "from the last 5 weeks", "this week").
4. Catch up → Copilot Chat. "Catch me up", "what did I miss".
5. A single question or quick lookup → Copilot Chat. "How do I…?", "Why am I seeing…?", "When is…?", "What is…?", "Who is…?", "Where…?" — including calendar, contact, and people lookups ("when's my next meeting?", "who's the PM for Contoso?", "what's on my calendar Friday?"), plus any quick fact, definition, policy lookup, or brainstorm. Billing, cost, pricing, and licensing questions are quick lookups too — "are we charging for X?", "how much does X cost?", "is X included in our license?", "do we pay for Y?" — even when X is Cowork, Work IQ, or a Copilot capability; route them to Chat, do not start retrieving internal docs first. Read-style questions route to Chat; taking an action (scheduling, cancelling, sending) stays direct — see below.
6. One document → Copilot in Word. "Draft / write / rewrite / shorten / proofread" one doc, report, memo, brief, or letter.
7. One workbook → Copilot in Excel. "Add a formula / column", "flag orders over $10k", "clean this CSV", "chart this" — within one file.
8. One deck → Copilot in PowerPoint. "Turn these notes into slides", "build a few slides", "suggest a layout".
9. Deep research → Researcher (included, subject to usage limits). "Research our competitors / this company / this market", "write a cited report".
10. Data analysis → Analyst (included, subject to usage limits). "Analyze this dataset", "find trends / anomalies", "run the numbers".
11. Domain workflow → specialized 1P agent (Sales / Service / Finance) — licensing varies; confirm entitlement.

When NOT to Use — not a routing decision

• Genuinely multi-step, multi-output projects (Cowork is the right tool — proceed, do not block): "research 3 competitors, build a comparison table, write a briefing, and turn it into a deck"; "from this one input, produce an Excel analysis + a written summary + a slide deck"; anything that chains several sources/steps into 2+ finished deliverables.
• Direct M365 actions (just do them — no free-vs-paid routing): "send / reply / forward an email", "schedule / cancel / decline / reschedule a meeting", "post to a Teams channel", "create a calendar event".
• Managing skills (creating, editing, validating, or auditing a skill) — use the skills skill.
• Pure execution the user already confirmed should run in Cowork.

Routing Logic — match the task to the right tool

Rule of thumb (from the field guidance): start simple and step up only if the task needs more. Chat → Copilot-in-app → Cowork.

Core Instructions

Phase 1 — Classify the task

Identify the capability bucket using the routing table. Ask: how many distinct finished outputs? how many apps/files? does it need web + work sources chained? One output / one app / one question → a free 1P capability almost always fits. Multiple outputs or many chained steps across sources → Cowork is legitimately the right tool.

Phase 2 — Recommend the free 1P capability

State the recommended capability and that it is included in the M365 Copilot license at no extra cost . Give a one-line "how to use it" (e.g. "open the file and use Edit with Copilot", or "ask Copilot Chat directly"). Keep it to 2–3 sentences. The goal is to train the user to go straight to the free 1P agent next time.

Phase 3 — Ask the user to confirm (required gate)

Use AskUserQuestion with two clear options:

• "Use [recommended 1P agent] — included" (recommend this)
• "Do it here in Cowork instead"

Do not proceed in Cowork on assumption. If the user picks the free agent, hand off with one-line instructions and stop. When that agent is M365 Copilot Chat, end the hand-off (toward the bottom of the message) with a clickable open link: [Microsoft 365 Copilot Chat](https://m365.cloud.microsoft/chat/) — then show the exact prompt right after it so the user can paste it in one step. (Microsoft does not currently support pre-filling a prompt via URL, so the link opens Copilot Chat but the user still pastes/types the prompt; never promise auto-fill and do not append the query to the url) If the answer comes back empty (cancelled), acknowledge and ask what they'd like.

Phase 4 — Proceed only if confirmed

Run the task in Cowork only after the user explicitly chooses Cowork. If they chose a 1P capability, do not run it in Cowork.

Output — what the end user sees (keep it crisp)

Two short parts, then the confirm gate. No pricing, no credits, no cost estimates of any kind. Use these exact bold labels:

Classification: <one line — what kind of task this is>.

Recommendation: <free 1P agent> can do this at no extra cost — it's included in your Microsoft 365 Copilot license. <One line on how/where to use it.>

Then the AskUserQuestion confirm gate. That's it — no long essays, no pricing explainer. Run in Cowork only if the user chooses Cowork.

Guardrails

• Confirmation gate is mandatory — always confirm before running the task in Cowork; never assume. Proceed in Cowork only on explicit user choice.
• No pricing or cost figures — never show credit counts, dollar amounts, pay-as-you-go rates, tiers, or cost estimates. Position the 1P route simply as included at no extra cost; do not quote numbers for anything.
• Lead users to the free 1P agent — recommend the included capability first; the intent is to minimize Cowork spend and train users to use M365 Copilot for these tasks. Defer to a stated user preference if they still want Cowork.
• Don't block legitimate Cowork work — if the task is truly multi-output or multi-step, say so and let the user proceed without friction.
• No fabricated capabilities or entitlements — for specialized agents (Sales, Service, Finance), say licensing varies and ask the user to confirm access.
• Handle ambiguity — if it's unclear how many outputs or apps the task spans, ask one quick clarifying question, or default to the simplest capable tool and say so. If you can't tell which tool fits, ask rather than guess.

Run this — do not improvise

This capability's deterministic implementation is a RAPP single-file agent, linked beside this file as what_to_use_when_agent.py and embedded as the fenced Python below (sha256 1175d5dd657078d4…; 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 what_to_use_when_agent.py first:

python3 what_to_use_when_agent.py '{"key": "value"}'      # arguments as one JSON object
echo '{"key": "value"}' | python3 what_to_use_when_agent.py   # or on stdin
python3 what_to_use_when_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.

"""WhatToUseWhen -- Run this FIRST, before ANY thinking, retrieval, task creation, or file generation, for any single-output task a free M365 Copilot can do; if action-bias pushes you to start, STOP and run this first. Fires on: "summarize my email/inbox/unread", "summarize my meetings", "summarize my Teams messages" (any window, e.g. "last 5 weeks"); "catch me up" / "what did I miss"; a single question or lookup: "how do I", "why", "when is", "what is", "who is", "are we", "do we", "is it", any billing/pricing/licensing question, calendar/contact lookups; "draft a doc" or memo; "create a presentation", deck, or spreadsheet; "research competitors"; "analyze this data". GUIDE the user to the matching free 1P agent (Copilot Chat; Word/Excel/PowerPoint; Researcher; Analyst), not Cowork; use it only if they insist. Do NOT fire for DIRECT ACTIONS (send email, schedule a meeting, post to Teams, create an event), long-running or multi-output jobs, or managing skills.

Generated by the rapp skill from what-to-use-when. 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 = '## Overview\n\nCowork runs on usage-based consumption credits on top of the Microsoft 365\nCopilot license. Most everyday single-output tasks can be done instead by a 1P\nCopilot capability that is **included in the M365 Copilot license at no extra\ncost**. This skill's job is to **guide the user to that free capability** —\nclassify the request, recommend the right 1P agent, and only proceed in Cowork if\nthe user explicitly insists. The intent is to minimize Cowork spend and train\nusers to reach for M365 Copilot for these tasks.\n\n## Precedence — run this BEFORE acting (most important rule)\n\nThis gate runs **before any thinking, retrieval, `TaskCreate`, file generation, or\ninvoking a doc/deck/spreadsheet skill (pptx / docx / xlsx)**. For an eligible\nrequest it **takes priority over action-bias and gather-before-generate**: if you\ncatch yourself about to reason through, search, create a task, or start building a\ndeck/doc/sheet for a single-output request (e.g. "create a presentation"),\n**STOP** and run Phase 1–3 below (classify → recommend → confirm) first. Only\nretrieve data or build anything after the user explicitly chooses an option.\n\n**Exclusion check FIRST (before any classification).** If the user's message is a\n**direct M365 action** — an imperative to send / reply / forward an email,\nschedule / cancel / decline / reschedule a meeting, post to a Teams channel, or\ncreate a calendar event — this gate does **NOT** apply. Do not classify, do not\nshow the confirm card. Hand straight back to the normal action path and let it\nexecute. These actions are out of scope even though they produce a single output,\nbecause the user is *taking an action*, not *asking for content a free 1P agent\ncould generate instead*. Only after a message clears this exclusion do you\nproceed to Phase 1.\n\n## Quick Start\n\n```\nUser: "Cowork, summarize my Teams messages from the last 5 weeks."\n\nThe user sees a crisp, 2-part reply that points them to the free option:\n\n  Classification: This is a quick Teams catch-up — a single summary.\n\n  Recommendation: Microsoft 365 Copilot Chat can do this at no extra cost —\n  it's included in your Copilot license. Just ask it in Copilot Chat.\n\nThen an AskUserQuestion gate:\n  [Use Copilot Chat — included] / [Do it here in Cowork instead]\nRun in Cowork only if the user picks Cowork.\n```\n\n## When to Use — fire on ANY of these (each is a separate trigger)\n\nCheck this **BEFORE retrieving M365 data or running the task in Cowork.** The\nquoted phrasings are cues — generalize to the intent, don't pattern-match.\n\n1. **Summarize email → Copilot Chat.** "Summarize my email / inbox / unread",\n   "any emails I need to reply to".\n2. **Summarize meetings → Copilot Chat.** "Summarize my meetings", "recap the\n   standup", "what did I miss in the meeting".\n3. **Summarize Teams → Copilot Chat.** "Summarize my Teams messages / chats /\n   channels", over **any** window (e.g. "from the last 5 weeks", "this week").\n4. **Catch up → Copilot Chat.** "Catch me up", "what did I miss".\n5. **A single question or quick lookup → Copilot Chat.** "How do I…?", "Why am I\n   seeing…?", "When is…?", "What is…?", "Who is…?", "Where…?" — including\n   calendar, contact, and people lookups ("when's my next meeting?", "who's the\n   PM for Contoso?", "what's on my calendar Friday?"), plus any quick fact,\n   definition, policy lookup, or brainstorm. **Billing, cost, pricing, and\n   licensing questions are quick lookups too** — "are we charging for X?", "how\n   much does X cost?", "is X included in our license?", "do we pay for Y?" — even\n   when X is Cowork, Work IQ, or a Copilot capability; route them to Chat, do not\n   start retrieving internal docs first. Read-style questions route to Chat;\n   *taking* an action (scheduling, cancelling, sending) stays direct — see below.\n6. **One document → Copilot in Word.** "Draft / write / rewrite / shorten /\n   proofread" one doc, report, memo, brief, or letter.\n7. **One workbook → Copilot in Excel.** "Add a formula / column", "flag orders\n   over $10k", "clean this CSV", "chart this" — within one file.\n8. **One deck → Copilot in PowerPoint.** "Turn these notes into slides", "build\n   a few slides", "suggest a layout".\n9. **Deep research → Researcher** (included, subject to usage limits). "Research our competitors /\n   this company / this market", "write a cited report".\n10. **Data analysis → Analyst** (included, subject to usage limits). "Analyze this dataset", "find\n    trends / anomalies", "run the numbers".\n11. **Domain workflow → specialized 1P agent** (Sales / Service / Finance) —\n    licensing varies; confirm entitlement.\n\n## When NOT to Use — not a routing decision\n\n- **Genuinely multi-step, multi-output projects** (Cowork is the right tool —\n  proceed, do **not** block): "research 3 competitors, build a comparison table,\n  write a briefing, and turn it into a deck"; "from this one input, produce an\n  Excel analysis + a written summary + a slide deck"; anything that chains several\n  sources/steps into 2+ finished deliverables.\n- **Direct M365 actions** (just do them — no free-vs-paid routing): "send / reply\n  / forward an email", "schedule / cancel / decline / reschedule a meeting",\n  "post to a Teams channel", "create a calendar event".\n- **Managing skills** (creating, editing, validating, or auditing a skill) — use\n  the **skills** skill.\n- Pure execution the user already confirmed should run in Cowork.\n\n## Routing Logic — match the task to the right tool\n\n| If the task is… | Recommend | Why |\n|---|---|---|\n| Quick question, find, brainstorm, summarize email/meetings/Teams | **M365 Copilot Chat** — *included* | Fast, conversational, grounded in your content |\n| Draft or edit **one** document | **Copilot in Word** (Edit with Copilot) — *included* | Works inside the file, edits in place |\n| Numbers in **one** workbook / a few slides in **one** deck | **Copilot in Excel / PowerPoint** — *included* | Understands that specific file |\n| Deep, multi-source research → cited report | **Researcher** — *included, subject to usage limits* | Deep-research + your work data |\n| Analyze a dataset, find trends, run the numbers | **Analyst** — *included, subject to usage limits* | Data-scientist agent, runs Python, shows work |\n| Domain workflow (sales, service, finance) | **Specialized 1P agent** (e.g. Copilot for Sales) — *licensing varies* | Purpose-built; confirm entitlement |\n| **One input → several finished outputs**, or a **multi-step job across files + web** | **Cowork** | Chains steps & produces multiple deliverables |\n\nRule of thumb (from the field guidance): **start simple and step up only if the\ntask needs more.** Chat → Copilot-in-app → Cowork.\n\n## Core Instructions\n\n### Phase 1 — Classify the task\nIdentify the capability bucket using the routing table. Ask: how many distinct\n**finished outputs**? how many **apps/files**? does it need **web + work\nsources chained**? One output / one app / one question → a free 1P capability\nalmost always fits. Multiple outputs **or** many chained steps across sources →\nCowork is legitimately the right tool.\n\n### Phase 2 — Recommend the free 1P capability\nState the recommended capability and that it is **included in the M365 Copilot\nlicense at no extra cost** . Give a one-line "how to use it" (e.g. "open the file and use Edit with\nCopilot", or "ask Copilot Chat directly"). Keep it to 2–3 sentences. The goal is\nto **train the user to go straight to the free 1P agent next time.**\n\n### Phase 3 — Ask the user to confirm (required gate)\nUse **AskUserQuestion** with two clear options:\n- "Use **[recommended 1P agent]** — included" (recommend this)\n- "Do it here in **Cowork** instead"\n\nDo not proceed in Cowork on assumption. If the user picks the free agent, hand\noff with one-line instructions and stop. **When that agent is M365 Copilot Chat,\nend the hand-off (toward the bottom of the message) with a clickable open link:**\n`[Microsoft 365 Copilot Chat](https://m365.cloud.microsoft/chat/)` — then show the\nexact prompt right after it so the user can paste it in one step. (Microsoft does\nnot currently support pre-filling a prompt via URL, so the link opens Copilot Chat\nbut the user still pastes/types the prompt; never promise auto-fill and do not append the query to the url) If\nthe answer comes back empty (cancelled), acknowledge and ask what they'd like.\n\n### Phase 4 — Proceed only if confirmed\nRun the task in Cowork **only** after the user explicitly chooses Cowork. If they\nchose a 1P capability, do not run it in Cowork.\n\n## Output — what the end user sees (keep it crisp)\n\nTwo short parts, then the confirm gate. **No pricing, no credits, no cost\nestimates of any kind.** Use these exact bold labels:\n\n> **Classification:** <one line — what kind of task this is>.\n>\n> **Recommendation:** <free 1P agent> can do this at no extra cost — it's\n> included in your Microsoft 365 Copilot license. <One line on how/where to use\n> it.>\n\nThen the **AskUserQuestion** confirm gate. That's it — no long essays, no pricing\nexplainer. Run in Cowork only if the user chooses Cowork.\n\n## Guardrails\n\n- **Confirmation gate is mandatory** — always confirm before running the task in\n  Cowork; never assume. Proceed in Cowork only on explicit user choice.\n- **No pricing or cost figures** — never show credit counts, dollar amounts,\n  pay-as-you-go rates, tiers, or cost estimates. Position the 1P route simply as\n  **included at no extra cost**; do not quote numbers for anything.\n- **Lead users to the free 1P agent** — recommend the included capability first;\n  the intent is to minimize Cowork spend and train users to use M365 Copilot for\n  these tasks. Defer to a stated user preference if they still want Cowork.\n- **Don't block legitimate Cowork work** — if the task is truly multi-output or\n  multi-step, say so and let the user proceed without friction.\n- **No fabricated capabilities or entitlements** — for specialized agents (Sales,\n  Service, Finance), say licensing varies and ask the user to confirm access.\n- **Handle ambiguity** — if it's unclear how many outputs or apps the task spans,\n  ask one quick clarifying question, or default to the simplest capable tool and\n  say so. If you can't tell which tool fits, ask rather than guess.'

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


class WhatToUseWhenAgent(BasicAgent):
    def __init__(self):
        self.name = 'WhatToUseWhen'
        self.metadata = {
          "name": "WhatToUseWhen",
          "description": "Run this FIRST, before ANY thinking, retrieval, task creation, or file generation, for any single-output task a free M365 Copilot can do; if action-bias pushes you to start, STOP and run this first. Fires on: \"summarize my email/inbox/unread\", \"summarize my meetings\", \"summarize my Teams messages\" (any window, e.g. \"last 5 weeks\"); \"catch me up\" / \"what did I miss\"; a single question or lookup: \"how do I\", \"why\", \"when is\", \"what is\", \"who is\", \"are we\", \"do we\", \"is it\", any billing/pricing/licensing question, calendar/contact lookups; \"draft a doc\" or memo; \"create a presentation\", deck, or spreadsheet; \"research competitors\"; \"analyze this data\". GUIDE the user to the matching free 1P agent (Copilot Chat; Word/Excel/PowerPoint; Researcher; Analyst), not Cowork; use it only if they insist. Do NOT fire for DIRECT ACTIONS (send email, schedule a meeting, post to Teams, create an event), long-running or multi-output jobs, or managing skills.",
          "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 what_to_use_when_agent.py
    #     python3 what_to_use_when_agent.py '{"arg": "value"}'
    #     python3 what_to_use_when_agent.py --tool          # emit the JSON tool contract
    _a = sys.argv[1:]
    if _a and _a[0] == "--tool":
        print(json.dumps(WhatToUseWhenAgent().to_tool(), indent=2))
    else:
        _raw = _a[0] if _a else (sys.stdin.read().strip() or "{}")
        print(WhatToUseWhenAgent().perform(**json.loads(_raw)))

# rci-capsule:v1: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

…(truncated)