Generated banner for Dt Coaching Foundation
Generated logo for Dt Coaching Foundation

Dt Coaching Foundation

MicrosoftOfficialPublished by Microsoft

  • Skill
  • Role or persona
  • Agent tooling
  • MIT
  • Low risk

From the author

Design Thinking coaching foundation knowledge: coach identity and philosophy, quality and fidelity constraints, method sequencing, coaching state schema, and the canonical deck workflow
— the description by Microsoft

AutoGineer summary

Written by AutoGineer from the skill’s own files, not by the author

The dt-coaching-foundation skill serves as the core knowledge base for a Design Thinking coach agent. It establishes a consistent persona based on the Think, Speak, Empower philosophy, ensuring the agent acts as a collaborative guide rather than a task executor. The skill provides the logic for session state management, method sequencing across nine Design Thinking stages, and the optional canonical deck generation workflow. It enforces quality constraints and fidelity rules while providing a progressive hint engine to support users at varying levels of expertise.

What it does

  • Defines the coach's identity, interaction philosophy, and communication boundaries.
  • Manages session state persistence and recovery protocols.
  • Provides a progressive hint engine to escalate support based on user needs.
  • Governs the canonical deck and customer-card generation workflow.
  • Enforces quality and fidelity constraints for all generated artifacts.
  • Guides method transitions and non-linear iteration patterns.

When to use it

  • At the start of a Design Thinking coaching session to initialize the agent's persona.
  • During method transitions to ensure adherence to the nine-method sequence.
  • When the team requires guidance on artifact generation or canonical deck creation.
  • When the agent needs to assess user readiness or provide progressive hints.
  • To recover session state after an interruption.

Advantages

  • Ensures consistent coaching behavior across all Design Thinking methods.
  • Maintains user agency by prioritizing exploratory questions over direct answers.
  • Provides structured, repeatable workflows for artifact creation.
  • Supports adaptive coaching through a tiered hint system.
  • Enforces quality standards without imposing rigid, non-negotiable constraints.

Usage examples

“I am stuck on the current method, what should I do?”
The agent will assess the current method, apply the progressive hint engine to provide a level-appropriate suggestion, and end with a question to maintain user agency.
“Can we generate the canonical deck now?”
The agent will check if the project has sufficient validated content, generate the requested markdown artifacts, and then prompt the user to decide on customer-card generation.

Install

  1. Download the skill as a ZIP:

    https://mp-staging.autogineer.com/agent-skills/microsoft/dt-coaching-foundation/download
  2. Unzip it into the cross-client skills folder of your project:

    mkdir -p .agents/skills && unzip dt-coaching-foundation.zip -d .agents/skills

Documentation by Microsoft

DT Coaching Foundation — Skill Entry

This SKILL.md is the entrypoint for the Design Thinking coaching foundation knowledge.

The dt-coach agent loads this skill at session start and during a coaching session to ground coaching behavior in a stable identity, enforce quality and fidelity expectations, navigate method transitions, persist and recover session state, and run the opt-in canonical deck workflow. The foundation knowledge stays constant across all nine Design Thinking methods.

Foundation references

Load the reference that matches the current coaching moment.

ReferenceWhen to load
coaching-identity.mdAt session start and throughout — establishes coach identity, Think/Speak/Empower philosophy, boundaries, and the Progressive Hint Engine.
quality-constraints.mdBefore any artifact generation — enforces fidelity rules, anti-polish stance, and quality-by-space expectations.
method-sequencing.mdAt method transitions, space boundaries, and when the team asks what to do next — guides the nine-method sequence, transition protocol, next-method assessment, and non-linear iteration.
coaching-state.mdFor persistence and session recovery — defines the coaching state schema, update rules, and recovery protocol.
canonical-deck.mdFor opt-in canonical deck and customer-card generation — governs activation, offer points, and the customer-card branch.

Skill layout

  • SKILL.md — this file (skill entrypoint).

  • references/ — the DT coaching foundation knowledge documents.

    • coaching-identity.md — coach identity, philosophy, and interaction conventions.

    • quality-constraints.md — fidelity rules and quality standards across all methods.

    • method-sequencing.md — method transitions, space boundaries, next-method assessment, and iteration patterns.

    • coaching-state.md — coaching state schema, file conventions, and recovery protocol.

    • canonical-deck.md — opt-in canonical deck and customer-card workflow.

Files

references/coaching-state.md


title: 'DT Coaching State Protocol' description: Coaching state schema, file conventions, and session management protocol for tracking Design Thinking method progress across sessions.

This instruction defines the coaching state schema, file conventions, and session management protocol for Design Thinking projects. The state file tracks method progress across sessions and enables the coach to resume seamlessly.

State File Location

Store the coaching state file at .copilot-tracking/dt/{project-slug}/coaching-state.md.

  • {project-slug} is a kebab-case project identifier provided by the user (e.g., factory-floor-maintenance). DT method artifacts are scoped under .copilot-tracking/dt/{project-slug}/.

  • Create the directory when initializing a new coaching project.

  • One state file per project. Multiple concurrent projects each get their own directory.

State File Schema

# .copilot-tracking/dt/{project-slug}/coaching-state.md
project:
  name: "Human-readable project name"
  slug: "kebab-case-identifier"
  created: "YYYY-MM-DD"
  initial_request: "Original customer request verbatim"
  initial_classification: "frozen | fluid"

current:
  method: 1          # integer 1-9
  space: "problem"   # problem | solution | implementation
  phase: ""          # free-text label for step within current method
  disclaimerShownAt: null  # ISO 8601 timestamp; set once when the DT disclaimer is shown for this session

methods_completed: []  # list of integers, e.g. [1, 2, 3]

transition_log:
  - from_method: null
    to_method: 1
    rationale: "Project initialized"
    date: "YYYY-MM-DD"

hint_calibration:
  level: 1              # integer 1-4 matching Progressive Hint Engine levels
  pattern_notes: ""     # free-text observations about user's hint responsiveness

session_log:
  - date: "YYYY-MM-DD"
    method: 1
    summary: "Brief description of session work"

artifacts: []
  # - path: ".copilot-tracking/dt/{project-slug}/stakeholder-map.md"
  #   method: 1
  #   type: "stakeholder-map"

canonical_deck:
  snapshots: []
    # - method: 1
    #   date: "YYYY-MM-DD"
    #   entry_count: 0
    #   candidate_count: 0  # number of candidate entries before filtering
    #   fingerprint: ""    # content hash for staleness detection
  last_offered_at: null  # ISO 8601 timestamp; last time a snapshot was offered
  last_offered_response: null  # "accepted" | "declined" | null
  last_generated_at: null  # ISO 8601 timestamp; last time a snapshot was generated

customer_card_render:
  last_offered_at: null   # ISO 8601 timestamp; last time customer-card generation was offered
  last_offered_response: null  # "accepted" | "declined" | null
  last_generated_at: null  # ISO 8601 timestamp; last time customer-card content was generated
  decline_final: false    # boolean; true if user declined at Method 5 (no further offers)
  output_path: null       # relative path to last generated content directory
Field Definitions
Project Block
  • name: display name for the project, set during initialization.

  • slug: kebab-case identifier matching the directory name.

  • created: ISO 8601 date when the project was initialized.

  • initial_request: verbatim customer request captured at project start. Preserved as-is for comparison against discovered problem space.

  • initial_classification: frozen or fluid classification from Method 1 assessment.

Current Block
  • method: integer 1-9 indicating the active method.

  • space: derived from method number. Methods 1-3 map to problem, 4-6 to solution, 7-9 to implementation.

  • phase: free-text label describing the current step within the method (e.g., "stakeholder mapping", "interview planning", "theme clustering", "prototype testing").

  • disclaimerShownAt: ISO 8601 timestamp recording when the Design Thinking Coaching disclaimer was shown. null until shown; once set, never overwritten. See the disclaimer-language skill (Design Thinking Coaching section).

Hint Calibration
  • level: integer 1-4 indicating the current Progressive Hint Engine level for this team. Start at 1; increase when the team needs more direct guidance and decrease when the team demonstrates self-direction. Levels match the coaching identity's Progressive Hint Engine (Broad Direction, Contextual Focus, Specific Area, Direct Detail).

  • pattern_notes: free-text observations about the team's hint responsiveness, learning pace, and coaching style preferences. Updated as patterns emerge across sessions.

Methods Completed

List of method numbers the team has finished. A method is complete when the coach and team agree its outputs are sufficient to proceed. Once added, methods remain in this list even if they are revisited later.

Transition Log

Chronological record of method transitions. Each entry captures:

  • from_method: source method number (null for initial entry).

  • to_method: target method number.

  • rationale: brief explanation of why the transition occurred.

  • date: ISO 8601 date.

Non-linear iteration produces backward transitions (e.g., from Method 6 back to Method 2). These are normal and recorded with rationale.

Session Log

Chronological record of coaching sessions. Each entry captures:

  • date: ISO 8601 date.

  • method: active method during the session.

  • summary: brief description of work accomplished.

Artifacts

List of artifacts produced during coaching. Each entry captures:

  • path: relative path to the artifact from workspace root.

  • method: method number that produced the artifact.

  • type: artifact type descriptor (e.g., "stakeholder-map", "interview-notes", "synthesis-themes", "concept-sketch", "prototype-feedback").

Workflow-Specific Extension Blocks

Specialized DT workflows may extend the base state schema with additional top-level blocks. Preserve these blocks when updating coaching state files.

  • canonical_deck: snapshot history and cooldown state for canonical deck generation.

  • customer_card_render: cooldown state and output metadata for customer-card slide content derived from the canonical deck.

Canonical Deck Block
  • snapshots: list of snapshot records, one per canonical deck generation. Each entry records the method number, date, entry count, candidate count, and content fingerprint for staleness detection.

  • last_offered_at: ISO 8601 timestamp of the most recent canonical deck offer, whether accepted or declined.

  • last_offered_response: user's response to the most recent canonical deck offer: "accepted", "declined", or null if never offered.

  • last_generated_at: ISO 8601 timestamp of the most recent canonical deck generation.

Customer Card Render Block
  • last_offered_at: ISO 8601 timestamp of the most recent customer-card offer.

  • last_offered_response: user's response to the most recent offer: "accepted", "declined", or null if never offered.

  • last_generated_at: ISO 8601 timestamp of the most recent customer-card content generation.

  • decline_final: boolean indicating the user declined at the Method 5 checkpoint. When true, no further customer-card offers are made.

  • output_path: relative path to the last generated customer-card content.yaml directory.

State Management Rules

Initialization

Create the state file when DT Coach session initialization starts a new coaching project. Set current.method to 1, current.space to problem, and record the initial transition log entry.

Updates

Update the state file at these events:

  • Method transition (forward, backward, or lateral): update current block and append to transition_log. When the transition reflects that the current method is complete (the coach and team agree its outputs are sufficient to proceed), add the departing method to methods_completed if not already present.

  • Session start: append to session_log with current date and active method.

  • Artifact creation: append to artifacts list.

  • Phase change within a method: update current.phase.

  • Hint calibration shift: update hint_calibration.level when the team's responsiveness to hints changes. Record observations in hint_calibration.pattern_notes.

Write Failures

When a state write fails, report the failure and its cause, state that the recorded state is unchanged, and ask the team how to proceed. Do not retry the write through another path on your own, and do not describe a transition or update as complete when its write failed.

Space Derivation

Always derive current.space from current.method:

  • Methods 1-3: problem

  • Methods 4-6: solution

  • Methods 7-9: implementation

Do not set space independently of method.

Session Recovery Protocol

When resuming a coaching session:

  1. Read the state file at .copilot-tracking/dt/{project-slug}/coaching-state.md. When the slug is missing or ambiguous, list the projects under .copilot-tracking/dt/ with their last session dates and ask which one to resume; when no project exists, offer to start a new one.

  2. Verify the file parses as valid YAML and contains required fields (project, current, methods_completed, transition_log).

  3. Restore coaching context from current.method, current.space, and current.phase.

  4. Review the most recent transition_log and session_log entries to understand where the team left off.

  5. Check methods_completed to understand overall progress.

  6. Scan the artifacts list for available project artifacts to reference.

  7. Announce the resumed state to the user: current method, current phase, overall progress, the most recent transition rationale when it helps, and a brief summary of previous work.

  8. Ask the user to confirm the summary before resuming active coaching, then continue naturally from where the previous session ended rather than reciting method steps.

  9. Resume hint escalation at the level recorded in hint_calibration and recent session logs rather than resetting to Level 1.

If the state file is missing or corrupted, inform the user which fields are unreadable and offer to reinitialize from scratch or reconstruct state from existing artifacts in the project directory.

Project Directory Contents

The .copilot-tracking/dt/{project-slug}/ directory holds all project-specific artifacts alongside the state file:

  • coaching-state.md: coaching state (this schema).

  • Method outputs: stakeholder maps, interview notes, synthesis documents, concept descriptions, prototype feedback, testing results.

  • Naming convention: use descriptive kebab-case filenames prefixed with the method number (e.g., method-01-stakeholder-map.md, method-03-synthesis-themes.md).

  • All DT coaching artifacts are scoped to .copilot-tracking/dt/{project-slug}/. Never write DT artifacts directly under .copilot-tracking/dt/ without a project-slug directory.

Integration with Method Sequencing

The coaching state schema aligns with the method routing assessment flow used during method sequencing. The current.method field drives which dt-methods reference the agent reads on demand when a method becomes active. The transition_log provides the history that the method sequencing transition protocol references.

Requirements and compatibility

Written for: GitHub Copilot, Claude Code

Works with: Claude Code, Claude apps, OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI, Kiro, OpenCode, Goose, Amp, Roo Code, JetBrains Junie, Factory Droid, Devin (ex-Windsurf), Mistral Vibe, OpenHands, Letta

“Works with” means the tool reads the open Agent Skills format; a skill written for one tool may use features others lack.

Needs: Claude Code frontmatter

Security

What our scanner noticed· Low risk

  • Instructions that use the network

A static scan is not a guarantee. Skills are third-party content: review a skill before you let an agent use it.

Attribution and licence

Published by
Microsoft
Publisher
Microsoft (Official)
Source
microsoft/hve-core · .github/skills/design-thinking/dt-coaching-foundation · commit f425e94
Licence
MIT License (redistribution allowed)
  • LICENSE (added by AutoGineer from LICENSE)
  • ATTRIBUTION.md (written by AutoGineer)
Synced

AutoGineer is not affiliated with or endorsed by Microsoft.

Download ZIP

The ZIP holds the skill’s files, its licence files and an ATTRIBUTION.md. Download dt-coaching-foundation.zip