What does this skill do?

The Intent-Driven Development Skill transforms ambiguous engineering or product requirements into observable and verifiable acceptance criteria, preventing the specification from becoming a mere bureaucratic formality and ensuring that the result is testable. Examine the technical context, choose the level of detail based on risk, and draft numbered criteria that include an initial condition, a trigger, an expected result, prohibited effects, and a verification method.

Clarify Features
Define observable acceptance criteria before implementing a new feature or product change.
High-Impact Changes
Identify changes that affect security, authentication, persistent data, migrations, or external APIs.
Handoff Between Teams
Prepare a structured handoff for another team or AI agent with clear scope and criteria.
Review specifications
Review existing PRDs or specifications to identify gaps, vague terms, or unverifiable requirements.

Usage examples

🔍 Ambiguous change
Let's add a CSV export feature to the dashboard. Generate acceptance criteria using Quick Capture.
🔐 Auth Migration
We migrated authentication to OAuth. Prepare a comprehensive report that includes a risk assessment and rollback criteria.
🤝 Technical Handoff
Prepare a handoff for the backend team with verifiable acceptance criteria for the new payments API.
🔎 Audit PRD
Review this PRD and mark the requirements that are not directly observable or testable.

Features

Context Inspection Review the repository, documentation, and schemas to gather technical facts before asking questions.
Two levels of depth Quick Overview (3–7 criteria, low/moderate risk) or Full Report (security, migrations, changes between systems).
Observable Criteria for AC-NNN Each criterion includes an initial condition, a trigger, an expected outcome, prohibited effects, a verification method, and a priority.
Revision Management If a criterion fails due to technical limitations, it is marked as [revised], the scope is updated, and the revision number is incremented.
Lockout due to critical risk Implementation is blocked only if there is a material security risk, data loss, irreversible migration, or significant cost.

Frequently asked questions

No. This skill is methodological and does not require a connection to any external MCP. It works directly in Claude Code or Cowork.
Quick Capture is for low- or moderate-risk changes (3–7 criteria). The Full Report is used when security, migrations, or changes between systems are involved.
You must describe the initial condition, trigger, expected result, prohibited effects, and verification method. Avoid vague terms such as "correctly" or "certainly."
Yes, the skill is licensed under the MIT license. You can customize the depth levels, criteria templates, and priorities to fit your workflow.
Intent-Driven Development — Verifiable Acceptance Criteria with Claude AI

¿Prefieres escuchar el contenido? Genera la narración de audio con un clic.