psychology

chevp-ai-framework

Lifecycle Gates Guidelines Commands Agents Templates
looks_one Step 1 of 3

Context

Understand the system and the problem before proposing solutions.

No code. No feature plans. Read, verify, ask.

visibility

Goal

Understand the system, the problem, and the scope. Produce the foundational artifacts that all subsequent work builds on.

schedule

When

Always. No exceptions. Even trivial changes require reading CLAUDE.md and confirming scope.

input

Inputs

User requirement, existing code & architecture, CLAUDE.md, project documentation.

Mandatory Deliverables

Nine artifacts — in order. Numbers 2–4 form the uncertainty triplet.

1
Context-Plan (CTX)

Lightweight plan: what to read/verify, open questions, draft scope boundaries.

2
Problem Statement uncertainty triplet

Who has the problem? What does not work today? Why does it matter? Why now?

3
Hypotheses uncertainty triplet

What we think could be true and could solve the problem; each hypothesis carries a cheapest test and a kill criterion.

4
Risks uncertainty triplet

Where we might be wrong, top-3 expensive failure modes, counter-evidence we may be ignoring.

5
System Spec

What the system is, what it does, who it serves, which components it has.

6
Software Architecture

High-level architecture: layers, modules, communication paths, technology stack.

7
ADRs

Architecture Decision Records for all fundamental design decisions.

8
Context Inventory

Catalogue of existing artifacts: code, docs, schemas, conventions, dependencies.

9
Scope Confirmation

Only after 1–8: confirm scope with the human.

tips_and_updates Verify Once per Session: if deliverables already exist and have not changed, confirm in one sentence — do not re-read.

The Uncertainty Triplet

Context is where uncertainty is explicitly catalogued — not assumed away

2

Problem Statement

Concrete answers to: who has the problem, what fails today, why it matters, why now, what counts as "solved".

3

Hypotheses

≥2 testable hypotheses. Each one ships with the cheapest way to confirm or kill it.

4

Risks

≥3 risks rated. Top-3 expensive failure modes called out, plus counter-evidence the team may be ignoring.

Read the uncertainty-reduction guideline →

The 6 Decisions Toward G1

Each surfaced via AskUserQuestion — never as prose

  1. Confirm the Problem Statement — accept / edit / regenerate
  2. Confirm the Hypotheses — accept / edit / regenerate
  3. Confirm the Risks — accept / edit / regenerate
  4. Approve the System Spec & Architecture snapshot
  5. Confirm scope — what is in / out / uncertain
  6. Pass G1 → Exploration (approve / reject / revise)

Quality Gate G1

Every checkbox must be satisfied before moving to Exploration

check_circleContext-Plan exists and is confirmed by human
check_circleProblem Statement exists, all five questions answered concretely
check_circleHypotheses exist (≥2 with cheapest test + kill criterion)
check_circleRisks exist (≥3 rated, top-3 failure modes called out)
check_circleSystem Spec exists (what the system is, who it serves)
check_circleSoftware Architecture documented (layers, modules, communication)
check_circleADRs exist for fundamental decisions
check_circleExisting artifacts catalogued (Context Inventory)
check_circleAffected files/modules identified, dependencies known
check_circleAll open questions resolved with the human
check_circleEvidence block filled (hypothesis / result / reasoning)
check_circleGatekeeper-G1 verdict recorded
check_circleHuman has explicitly confirmed scope
block BLOCKER: Do not proceed to Exploration until every checkbox is satisfied.

Roles Active in This Step

Four of the seven cross-cutting roles operate in Context

SDLC

Process governance, scope confirmation, gate enforcement.

AI-Plans

System Spec, problem formulation, initial scope definition.

Software-Architecture

Architecture document, ADRs, codebase analysis.

Context-Engineering

Context Inventory, what AI must read, context hierarchy.

arrow_back Lifecycle overview Step 2: Exploration arrow_forward
menu_book

Read the source

01-context/README.md open_in_new