Digital Innovation Agents

Appearance

DIA Bootstrap ​

dia-bootstrap is the bootstrap skill (renamed from using-digital-innovation-agents in v3.3). It loads automatically at session start via the SessionStart hook (or the OpenCode plugin's transform) and gives the agent its introduction to the workflow on every new session. You do not invoke it manually.

The skill exists for two reasons:

  • Discoverability. A new user types nothing special and the agent already knows there is a V-Model available, what the entry points are, and how to opt out.
  • Boundary safety. The skill makes the workflow advisory. Without it, the agent might push the workflow even when the user just wants a quick edit. With it, the agent asks first and respects opt-out language.

What the agent learns at session start ​

The skill catalog ​

The agent learns it has access to thirteen skills, grouped into guide, brownfield entry, V-Model phase skills, and foundation skills. The full list is in Commands.

When to invoke which skill ​

User contextSuggested entry
Starting something new, problem space unclear/dia-guide
Existing codebase, no V-Model artifacts/reverse-engineering, then /business-analysis for the WHY
Existing DIA project on v1 or v2/dia-migration
Clear problem, no solution yet/business-analysis
Features defined, no architecture/architecture
Ready to implement/coding
A bug surfaced outside an active cycle/coding (bug-capture entry)

These are suggestions, not rules. The user is in charge.

Language adaptation ​

Skill content is written in English so it is portable. The agent always responds in the user's language. German user, German reply. Spanish user, Spanish reply. Skill instructions stay English internally; user-facing messages adapt automatically.

Artifact locations ​

Project artifacts live under _devprocess/:

  • _devprocess/analysis/ for BA, EXPLORE, RESEARCH, AUDIT, EPIC-NN-ba (flat, no subfolders except sources/)
  • _devprocess/requirements/ for epics, features, fixes, improvements, handoff
  • _devprocess/architecture/ for ADRs and arc42
  • _devprocess/rules/ for stable rule sets (technical, design, domain)
  • _devprocess/implementation/plans/ for PLAN-NN files
  • _devprocess/context/ for BACKLOG.md (single source of truth for state), HANDOFFS.md, METRICS.md
  • src/ARCHITECTURE.map for the wayfinder root
  • src/{module}/README.md for module wayfinders

See Artifacts for the full directory tree.

Opt-out behaviour ​

The skill set is advisory. The user can leave the workflow at any time, and the agent respects that immediately.

Leaving a /dia-guide loop mid-cycle ​

User says "stop", "exit", "I want to do something else", "let's pause this", or asks an unrelated question:

  • The agent exits the workflow immediately.
  • No "are you sure?" pushback.
  • Answers the next question directly, without invoking any V-Model skill.
  • Workflow state stays in _devprocess/ and in the git phase-done tags. The user can resume later by re-invoking /dia-guide.

Temporarily disabling ​

User says "ignore V-Model today", "I just want a quick fix", "no skills needed for this", "just help me with X without the workflow":

  • The agent does not invoke any skill, even when the task matches a skill description ("fix this bug" would normally match /coding).
  • Plain mode, like any normal Claude Code session without the plugin.
  • The agent does not remind the user that the skills exist or suggest re-enabling.
  • The opt-out stays in effect until the user explicitly ends it or a new session starts.

Permanently disabling ​

In the Claude Code CLI:

/plugin disable digital-innovation-agents

This removes the SessionStart hook entirely. The skill mentions this command only when the user asks how to disable permanently.

User Interaction Protocol ​

The orientation skill teaches every phase skill the same dialog rules:

  1. One question per turn. Never batch decisions.
  2. Use AskUserQuestion for choices. Free-form prose only for quick factual confirmations.
  3. Each option carries an explicit Pro and Con line. The user sees both sides at a glance.
  4. Mark the recommended option as the first entry with "(Recommended)". If the rationale is not obvious from the Pros / Cons, the agent adds a one-line "Empfehlung: ... weil ..." sentence before the question.
  5. No "dealer's choice" framing. If the agent has no preference, it says so explicitly.

These rules bind across every phase skill, regardless of project language.

Principles ​

  • Living documents. Every phase writes back into its source artifacts so documentation reflects the current state. State goes through the backlog row first.
  • Tech-agnostic success criteria. No OAuth, REST, or PostgreSQL in Success Criteria. Technology details belong in Technical NFRs.
  • Quality gates. Each skill verifies its own output before handoff. /consistency-check Mode A runs at every phase boundary.
  • User in control. No autonomous generation, always propose and confirm.
  • Advisory, not enforcing. If the user doesn't want the workflow, the agent doesn't force it.

Scope adaptation ​

The workflow adapts to project scope:

  • Simple Test / Feature (hours to 1 to 2 days): minimal exploration, skip validation, focus on definition of done.
  • Proof of Concept (1 to 4 weeks): shortened exploration, full ideation, hypothesis-driven validation.
  • Minimum Viable Product (2 to 6 months): full exploration, full ideation, complete market assessment.

/business-analysis asks which scope applies at the start and calibrates depth accordingly.

Read the skill file ​

skills/dia-bootstrap/SKILL.md on GitHub.

See also ​