title	Intent
type	framework
maturity	final
confidentiality	shareable
created	2026-03-29
updated	2026-07-02
thought_leaders	marty-cagan jeff-patton teresa-torres josh-seiden
frameworks	product-operating-model outcomes-over-output
version	0.11.0
architecture	sibling-composable
pipeline_position	team-operating-model
port_contract	consumes signals from practitioner surfaces; produces specs + contracts + events consumable by execute-layer tooling (Spec Kit, Kiro, Copilot, Claude Code)

Intent — v0.11.0

A team operating model for AI-augmented product teams.

When AI collapses implementation from weeks to hours, the bottleneck moves upstream — from delivery to discovery, specification, and observation. Agile ceremonies built for slow, expensive delivery become overhead. The team's operating model needs to change.

Intent is that operating model.

---

Start here

Want the plain-language version, no jargon? → translation/intent-in-plain-language.md

Run it:

make setup   # builds the servers venv (fastmcp, pyyaml, pytest)
make test    # 169 repo-local tests, all green

Reading path (for engineers — read in this order):

1. servers/knowledge.py + servers/test_knowledge.py — the knowledge/coherence engine and its 56-test suite (the audit_chain, lint, query, and entity-resolution logic).
2. spec/SPEC-INTENT-COHERENCE-GATE-001.md — the load-bearing design: why the synthesis step is a coherence gate, not a merge.
3. knowledge/decisions/DDR-009-externally-authored-verification.md — a worked decision record: provider/model-separated verification.
4. ARCHITECTURE.md — how the pieces fit.

What's code vs. what's working memory: servers/, tools/, bin/, and formation/ are the running system. .intent/ (180+ signals), knowledge/, and spec/ are the dogfood — Intent tracking its own development through its own loop. It's a working record, not product surface; browse it for evidence of the method, not as an API.

Two products share this repo. Intent (this methodology) and the Knowledge Engine (knowledge-engine/, a separable compiled-knowledge product) are distinct. See CLAUDE.md §"This Repo Contains Two Products" before diving in.

---

The Core Loop

NOTICE  →  SPEC  →  EXECUTE  →  OBSERVE  →  (back to NOTICE)

No sprint boundaries. No ceremony tax. A continuous loop where the team's energy follows the highest-leverage work:

Phase	What Happens	Replaces
Notice	Sense signals — user feedback, system patterns, market shifts	Backlog grooming
Spec	Shape into Intent → Shape → Contract (the full team collaborates)	Sprint planning + ticket writing
Execute	Humans + AI agents build against the spec	Sprint execution
Observe	Verify outcomes against reality, feed learnings back	Sprint review + retro

---

Three-Layer Architecture

Intent prescribes three independently useful layers:

Layer	Purpose
Compiled Knowledge Base	Raw sources compiled into durable domain knowledge (personas, journeys, DDRs, models)
Transformation OS	The Notice→Spec→Execute→Observe engine with trust scoring, autonomy levels, and event stream
Software Spec & Code	Agent-ready specs, verifiable contracts, and running implementation

Six bidirectional flows couple the layers. The double-loop (Observe → Knowledge) questions assumptions, not just optimizes.

Full architecture: ARCHITECTURE.md

---

This Repo Eats Its Own Dogfood

The directory structure is the loop:

.intent/         ← Signals, intents, specs, contracts, events (180+ signals captured)
spec/            ← Shaped specifications: methodology, trust framework, enforcement specs
learnings/       ← IDD playbook, cross-product applicability, process drift catalog
hooks/           ← 8 governance hooks (autonomy-grant, closure-discipline, overwatch)
spawn-prompts/   ← 8 reusable handoff prompts for governance and build operations

---

Where Intent Sits

Intent fills a distinct pipeline position — one not currently occupied by any existing product category. The table below renders a signal-propagation ordering (each position produces artifacts the adjacent position reads), not a containment hierarchy.

Position	What It Solves	Examples (siblings at this position)
Team Operating Model	How the team flows, decides, and learns	Intent ← this (siblings: Shape Up team model, Scrum, bespoke team OS)
Spec-Driven Development	How specs become code via AI agents	GitHub Spec Kit, Kiro, Tessl
AI Coding Assistants	How individuals get AI help	Copilot, Claude Code, Cursor

Intent is the currently-unfilled position in that pipeline. It is not "above" the other two — it composes with them at declared handoffs. See Workspaces/.context/DECISIONS.md (WS-DDR-025) for the sibling-over-parent-child default.

---

Governance Infrastructure (v0.11.0)

Intent's build-discipline layer is operational as of v0.11.0:

IDD Build Discipline — 7-stage loop with DoR/DoD gates. Canonical playbook: learnings/idd-playbook.md. Cross-product map: learnings/idd-cross-product-applicability.md.

Process Drift Catalog — 4 families, 17 entries covering every recurring AI-agent drift pattern observed across the Intent system. learnings/process-drift-catalog.md. New in v0.11.0: 1.7 — Artificial-gate-architecture drift and 4.7 — Governance-skill-without-trigger.

5-Layer Autonomy-Grant Enforcement — Prevents L4-eligible work from being converted to L0 proposals. Spec: spec/autonomy-grant-enforcement.md. Hooks: autonomy-grant-check.sh (SessionStart), autonomy-grant-stop-check.sh (Stop), autonomy-grant-dispatch-prompt-check.sh (PreToolUse).

5-Layer Closure-Discipline Enforcement — Requires upstream control before status: resolved. Spec: spec/closure-discipline-enforcement.md. Hooks: closure-discipline-check.sh (SessionStart), closure-discipline-stop-check.sh (Stop), closure-discipline-signal-check.sh (PreToolUse).

Overwatch Staleness Hook — SessionStart banner when latest overwatch journal is >7 days (warn) or >14 days (load-bearing posture). hooks/overwatch-staleness-check.sh.

---

Interactive Artifacts

• Intent Visual Brief — Seven-section interactive product brief (Problem cascade, Core Loop, Stack, Personas, GTM, Hypotheses, Competitive)
• Intent-Native Repos — How to adopt Intent across repos: template structure, adoption tiers, and the Entire.io → Spec feedback loop

Documentation & marketing site: theparlor.github.io/intent-site (separate repo: theparlor/intent-site)

---

Status

Version: 0.11.0 (2026-05-20) Stage: Idea → ready for validation Approach: Thought leadership first, methodology product second, tooling conditional on validation Governance infrastructure: Operational — 8 hooks, 8 spawn prompts, IDD playbook, 4-family drift catalog Next step: 5 in-depth interviews with teams experiencing AI + Agile friction

---

Origin

Designed in a Cowork session on 2026-03-28, triggered by a conversation with engineer Ari about how AI collapses Agile Scrum. Grounded in Cagan (product operating model), Patton (story mapping), Torres (continuous discovery), and Seiden (outcomes over outputs).

---

Private repo. Will open when the manifesto is ready.