Architect a new runtime

Kernel hydration plan — which modules this session loads, in what order, under which budgets and mode. Deterministic and fail-closed: the token-minimal set. Loads nothing; every future runtime inherits this lifecycle.

architectlargehydrate15 hydrated4 excludedcheckpointgov-syncloads: nothing#afb48eca

Mode governance (referenced, not restated)

modearchitect

read strategyAI_CONSTITUTION + the seams a new subsystem consumes; MASTER_* only here

validationdesign only — no implementation unless separately approved

hydrationconstitution + consumed seams + roadmap pointer

token philosophydeep reasoning licensed; still no prose dumps

→ .ai/ENGINEERING_MODES.md

Boot order (deterministic)

session-boot → engineering-profile → execution-contract → token-policy → repository-memory → advisor → workspace-state → affected-runtime

Budgets (this tier)

context.ai/TOKEN_POLICY.md

template runtime + direct deps + governance heads

token.ai/TOKEN_POLICY.md

milestone budget; reuse over creation; functional-tie → fewer tokens

reasoning.ai/ENGINEERING_MODES.md

plan first, then build

memory.ai/SESSION_COMPRESSION.md

checkpoint + compress on long sessions

Module decisions (fail-closed: default is to exclude)

session-boothydrate

always_loaded · session-boot is part of the always-loaded boot chain

→ .ai/SESSION_BOOT.md

engineering-profilehydrate

always_loaded · engineering-profile is part of the always-loaded boot chain

→ .ai/ENGINEERING_PROFILE.md

execution-contracthydrate

always_loaded · execution-contract is part of the always-loaded boot chain

→ .ai/EXECUTION_CONTRACT.md

token-policyhydrate

always_loaded · token-policy is part of the always-loaded boot chain

→ .ai/TOKEN_POLICY.md

repository-memoryhydrate

always_loaded · repository-memory is part of the always-loaded boot chain

→ .ai/REPOSITORY_MEMORY.md

repository-snapshotskip

lazy_not_required · repository-snapshot not required by this task (lazy)

→ .ai/CONTEXT_LOADING_RULES.md

advisorhydrate

milestone_requires · advisor required by "milestone"

→ .ai/CONTEXT_LOADING_RULES.md

workspace-statehydrate

milestone_requires · workspace-state required by "milestone"

→ .ai/WORKSPACE_STATE.md

affected-runtimehydrate

task_requires · affected-runtime required by "implementing"

→ .ai/CONTEXT_LOADING_RULES.md

ai-constitutionhydrate

mode_requires · ai-constitution required by "architecting"

→ .ai/AI_CONSTITUTION.md

session-compressionskip

lazy_not_required · session-compression not required by this task (lazy)

→ .ai/SESSION_COMPRESSION.md

session-endhydrate

lifecycle_resolved · session-end required by "shutdown"

→ .ai/SESSION_END.md

release-policyhydrate

lifecycle_resolved · release-policy required by "shutdown"

→ .ai/RELEASE_POLICY.md

master-roadmaphydrate

mode_requires · master-roadmap resolved as a reference pointer (architecting) — not loaded in full

→ MASTER_ROADMAP.md

manifestohydrate

mode_requires · manifesto resolved as a reference pointer (architecting) — not loaded in full

→ MANIFESTO.md

decisionshydrate

mode_requires · decisions resolved as a reference pointer (architecting) — not loaded in full

→ .ai/DECISIONS.md

changelog-logskip

deferred_not_required · changelog-log not required by this task (deferred)

→ .ai/CHANGELOG_AI.md

master-architecturehydrate

mode_requires · master-architecture required by "architecting"

→ MASTER_ARCHITECTURE.md

execution-lifecycleskip

future_not_built · execution-lifecycle is a future module — never hydrated until built

→ MASTER_ROADMAP.md

Request inputs

task tierlarge

modearchitect

operationarchitect

needs machine statefalse

is milestonetrue

lifecycle phasehydrate

← all kernel plans