Mission Atlas

A Mission Atlas is a semantic map of many IntentLang missions. It lets a team understand a large change without opening every .intent file.

One mission is a single unit of intent. Real work produces many of them: one per feature, one per user journey, one per product area, one per MVP, one per release, or one full day of Claude Code or Codex output. Reading them file by file does not scale. The Atlas is how IntentLang scales beyond one file.

For the full family of scaling concepts (chains, digests, proof matrix, risk radar, semantic diff, MVP readiness), see Working with large changes.

The hierarchy

A Mission Atlas organizes intent from the product down to the code evidence:

Product or MVP
  User journeys
    Feature areas
      Mission chains
        Missions
          Guarantees
            Tests
              Proof
                Code evidence

Each level answers a different question:

• Product / MVP , what are we shipping?
• User journeys , what can a user actually do end to end?
• Feature areas , how is the work grouped (Identity, Billing, ...)?
• Mission chains , which missions form one flow? (see Mission chains)
• Missions , what is each single unit of intent?
• Guarantees / never rules , what must always or never be true?
• Tests / proof / code evidence , is it actually verified in the repo?

The Atlas is not a file tree. It is a meaning tree. Two missions in different folders can belong to the same chain; a single feature area can span many folders.

Mission Capsule

Opening a mission to understand it is slow. A Mission Capsule is a compact summary of one mission that answers the review questions at a glance:

• What is it?
• Why does it matter?
• What is risky?
• Is it verified?
• What changed?
• Should I review it?

Example capsule:

ResetPassword
Purpose:       Let users recover access safely.
Risk:          High security
Proof:         Partial
Tests:         6 generated, 4 passing
Never rules:   3, 1 unverified
Changed today: yes
AI authored:   yes
Human reviewed: no

A capsule is derived, not authored. It is computed from the mission's declared guarantees, never rules, and verify blocks, plus repo evidence (which tests exist, which pass) and change metadata (what moved, who authored it, whether a human reviewed it). The Atlas is a tree of capsules.

What the Atlas is for

• Navigate a product area without reading every mission.
• Locate the risky missions instead of reviewing 200 equally.
• See whether an end-to-end journey is actually complete, not just whether individual missions parse.
• Brief a reviewer, a new engineer, or an AI agent with a trustworthy map.

Teaching line: do not make people read 200 missions. Make the system answer what exists, what changed, what is risky, what is verified, and what blocks deployment.

Where the Atlas comes from (planned commands)

This repo teaches the Atlas as a concept, ships a worked example (examples/mvp-customer-portal/), and ships the first real indexing command. The rendered Atlas view and the interactive dashboard are owned by the SkillsTech Compiler and SkillsTech.

intent index is real today. It reports only what is derivable from the .intent files themselves: mission, feature area, a risk heuristic, guarantee and never counts, and declared verification (whether verify tests exist). It deliberately does not report test pass counts, proof, or drift, because those need a test runner and OpenThunder. mission-index.json in the example is generated by this command; the richer capsule fields above (tests passing, drift, human review) come from repo evidence and are still taught through the other example fixtures. See the customer-portal example and Working with large changes.

Boundaries

• IntentLang (this repo) teaches the concept and the examples.
• The SkillsTech Compiler owns indexing and the machine-readable artifacts.
• SkillsTech owns the product dashboard.
• OpenThunder verifies code against missions (drift).
• Repo Mastery teaches repo ownership; SkillsTech Certified builds certification material.

The Atlas never claims code is correct. It maps declared intent and reports what the repo can prove. Correctness against the running code is OpenThunder's job.