On this page
References
Index
Use this page as the map once the overview is not enough.
Cards follow the same order as the left sidebar, so the index stays predictable.
Overview
StartThe splash, first-command promise, visible proof cards, and path into the portal.
Plain EnglishWhat to paste, what gets written, what failures do, and what the terms mean.
What it isThe on-disk contract and the claims IMPLEMENTAUDIT does not make.
WorkflowThe request-to-proof path, nested loops, and failure interrupt behavior.
Success and proofProof cards, closure markers, and source-backed review surfaces.
First run
Quick StartThe first command, work record, run folder when phased, checks, failure path, and success markers.
InstallationClaude Code and Codex install paths, verification caveats, and update boundaries.
Possible endingsVerified complete, audit-gap repair loop, blocked handoff, and unsafe/unbounded stop.
ComparisonHow IMPLEMENTAUDIT differs from a normal agent run and from audit-only review.
Audience
For new usersA slower explanation of the Quick Start path for readers who want less jargon.
For agents and operatorsHow an agent should preserve run state, owner/source, checks, and handoff evidence.
For auditors and maintainersHow to inspect closure evidence without accepting chat-only claims.
Core model
Runtime modelConcentric loops, Andon interrupt, final audit loop, and markers.
Audit gate modelNon-skippable gates: safety, input, repo inspection, before/after checks, implementation, final audit, and trace.
Invocation shapesDirect governance, embedded governance, goal synthesis, and governed casual-build intake.
Usage examplesGood commands, governed casual-build intake, goal synthesis, bad examples, and worked traces.
Planning and phasesPlanner depth, phase specs, owner/source, checks, and phase closure markers.
RoutingExisting-repair route, new-design route, mixed work, and optional sidecar boundaries.
Operating methodLean terms as behavior: check, stop, learn, and update the standard only when warranted.
Evidence
State and artifactsWork record, run folder when phased, state file, phase checklist, evidence, logs, and final audit ledger.
Repo state comparisonDeliverable changes versus run-root evidence and baseline comparisons.
Error handlingStop, classify, inspect cause, countermeasure, rerun, escalate, or hand off.
Evidence boundariesWhat the docs can prove and what remains an owner or host check.
Optional toolingQualified Graphify terrain and ActiveGraph checkpoint/mirror boundaries.
Child-agent review loopsSpecialist review roles as evidence inputs, not authority.
Closure
Completion semanticsWhat
AUDIT_COMPLETE, handoff, and run completion mean.
Continuity and resumeHow the run resumes from files without trusting chat memory.
Repository
Repo layoutWhere docs, tests, scripts, references, templates, fixtures, and generated output live.
Package contentsWhat ships in the skill payload and what stays repo-side.
Repo recordRelease evidence, changelog entries, audit ledgers, and current-status boundaries.
Research & Engineering Lineage
OverviewWhy the public vocabulary is visible, how evidence is bounded, and why properties map to native controls rather than methodology modes.
LeanTPS and Lean genealogy, surviving engineering properties, criticism, and rejected toolkit universals.
AgilePre-Manifesto ancestry, Scrum and XP distinctions, later hybrids, empirical limits, and ceremony stripping.
WaterfallRoyce and the caricature, distinct V-shaped families, and the assurance properties that survive fixed phases.
Evolved-LAWThe cross-method property synthesis, adoption classifications, native RXX owners, and explicit non-claims.