# Prysai LLM Playbook: Table of Contents v0.2 Readers who want the platform-neutral method first can open the [Universal LLM first-task route](routes/universal-core-foundations-EN.md). It begins with one fictional, text-only rewrite before mapping four extracted units. It remains `candidate / not_run`; the route does not establish cross-platform behavior, learner outcomes, or completion time. > English source-locale directory page. The legacy unsuffixed draft was > removed during release cleanup; `book/table-of-contents-EN.md` is the > canonical English table of contents. It keeps all 22 chapters, > 18 experiments, maturity boundaries, and real-user problem research > entrances visible without claiming that the chapter or lab corpus is already > translated or runtime-verified. ## Migration state and link policy - Page `content_status`: `candidate`; source revision: `dd08a68`. - The directory contains 22 chapter records and 18 real experiment files. - Chapters are `candidate`. Labs are `draft` with `run_status: not_run`. - Chapter 6 has `claim_status: disputed`; Chapter 22 has `claim_status: current | disputed`. - English chapter sources are now available for chapters 1–22 and are the canonical English reader route. Labs remain a separate migration track: each lab keeps its own `draft` and `not_run` state, and a pending label means the English lab file has not yet been reviewed into the locale matrix. Shared governance, evaluation, and research documents are marked `locale-neutral`. - No link on this page silently falls back to another locale. A migration notice is part of every link whose localized target does not yet exist. ## Start here - [Chapter 0 — What Is a Large Language Model](guides/llm-fundamentals-EN.md) - [Chapter 1 — Understand GPT before Codex](chapters/01-gpt-and-codex-EN.md) - [Chapter 2 — Complete a safe, verifiable task](chapters/02-first-safe-task-EN.md) - [Project entry — EN](../README-EN.md) - [Book guide — EN](README-EN.md) - [Project map — EN](../docs/project-map-EN.md) - [Canonical chapter navigation source](../docs/governance/book-navigation.yaml) - [Preface — EN](preface-EN.md) - [Learning path contract — locale-neutral](../docs/governance/learning-path.yaml) - [Locale matrix — locale-neutral](../docs/governance/locale-matrix.yaml) - [Universal LLM first task — EN, optional application after Chapter 0 through Chapter 2](routes/universal-core-foundations-EN.md) - [First Safe Change — optional offline fixture before Lab 001](routes/first-safe-change-EN.md) ## Part I: From understanding GPT to the first safe use ### Chapter 1 — Understand GPT before understanding how Codex works How models generate from context; how Codex connects a model to a work environment; and how context, tools, Skills, permissions, and an observable Agent loop affect the result. **content_status:** `candidate` - Chapter: [Chapter 1 — EN source](chapters/01-gpt-and-codex-EN.md) - Experiment: [Lab 011 — EN source](labs/lab-011-gpt-codex-boundaries-EN.md) ### Chapter 2 — Complete a first safe, verifiable task Choose a low-risk task, write a first-task protocol, set confirmation points, and leave delivery evidence. **content_status:** `candidate` - Chapter: [Chapter 2 EN source](chapters/02-first-safe-task-EN.md) - Novice fixture: [First Safe Change — EN source](routes/first-safe-change-EN.md) (`candidate / not_run`; not a Lab run) - Experiment: [lab-001 EN source](labs/lab-001-first-safe-task-EN.md) ### Chapter 3 — Turn a wish into a task protocol Define the goal, background, inputs, constraints, allowed actions, acceptance, failure handling, and delivery format. **content_status:** `candidate` - Chapter: [Chapter 3 — EN source](chapters/03-task-protocol-EN.md) - Experiment: [Lab 002 — EN source](labs/lab-002-task-protocol-EN.md) ### Chapter 4 — Context, permissions, and the Agent action boundary Context layers, trust boundaries, sandboxing, approvals, external side effects, and observable behavior. **content_status:** `candidate` - Chapter: [Chapter 4 — EN source](chapters/04-context-permissions-and-agent-EN.md) - Experiment: [Lab 007 — EN source](labs/lab-007-action-boundaries-EN.md) ### Chapter 5 — Choose the right Codex surface How to choose among Local, Worktree, and Cloud, then select the desktop app, CLI, IDE, or web entry point for a task. **content_status:** `candidate` - Chapter: [Chapter 5 — EN source](chapters/05-choose-the-codex-surface-EN.md) - Experiment: [Lab 007 — EN source](labs/lab-007-action-boundaries-EN.md) ### Chapter 6 — Model selection is not model worship Compare models with a task set, cost, speed, stability, and verification; test assumptions about model positioning. **content_status:** `candidate` · related volatile claim: `claim_status: disputed` - Chapter: [Chapter 6 — EN source](chapters/06-model-selection-EN.md) - Research: [OpenAI/Codex baseline — locale-neutral research](../docs/research/openai-codex-baseline.md) ## Part II: From user to workflow designer ### Chapter 7 — How Skills, Plugins, MCP, and tools divide the work Understand the method, connection, execution, and distribution layers; choose the smallest effective capability combination. **content_status:** `candidate` - Chapter: [English source](chapters/07-skills-plugins-and-tools-EN.md) - Experiment: [Lab 004 — EN source](labs/lab-004-skill-selection-EN.md) ### Chapter 8 — The complete lifecycle from definition to delivery Definition, planning, building, verification, review, delivery, and maintenance, with vertical slices that remain verifiable. **content_status:** `candidate` - Chapter: [English source](chapters/08-full-lifecycle-workflow-EN.md) - Primary experiment: [Lab 013 — EN source](labs/lab-013-l3-vertical-slice-EN.md) - Supporting experiment: [Lab 009 — EN source](labs/lab-009-engineering-lifecycle-EN.md) ### Chapter 9 — Verification, doubt, and recovery Break completion claims into claims and evidence; handle uncertainty, failure, and recovery. **content_status:** `candidate` - Chapter: [English source](chapters/09-verification-and-recovery-EN.md) - Experiment: [Lab 003 — EN source](labs/lab-003-evidence-review-EN.md) ### Chapter 10 — Planning and vertical slicing Break a large goal into delivery slices with clear dependencies, runnable steps, and inspectable checks. **content_status:** `candidate` - Chapter: [English source](chapters/10-planning-and-slicing-EN.md) - Experiments: [Lab 002 — EN source](labs/lab-002-task-protocol-EN.md) · [Lab 013 — EN source](labs/lab-013-l3-vertical-slice-EN.md) ### Chapter 11 — Design a genuinely useful Skill Trigger boundaries, progressive disclosure, resources, scripts, outputs, failure examples, evaluation, and versioning. **content_status:** `candidate` - Chapter: [English source](chapters/11-designing-a-skill-EN.md) - Experiment: [Lab 005 — EN source](labs/lab-005-design-a-skill-EN.md) ### Chapter 12 — The Agent loop, state, and stopping conditions Observation, planning, action, feedback, retry, confirmation, and stopping; explain behavior without inventing hidden reasoning. **content_status:** `candidate` - Chapter: [English source](chapters/12-agent-loop-and-stop-EN.md) - Experiment: [lab-006 — English source](labs/lab-006-agent-stop-conditions-EN.md) ### Chapter 13 — Action boundaries for files, terminals, browsers, and GitHub Read-only checks, editing, commands, browsing, commits, pushes, external messages, and rollback. **content_status:** `candidate` - Chapter: [English source](chapters/13-action-boundaries-EN.md) - Experiment: [Lab 007 — EN source](labs/lab-007-action-boundaries-EN.md) ## Part III: Skills, tools, and professional practice ### Chapter 14 — Discover, install, and audit external Skills Move from an index to a trustworthy capability: source, license, dependencies, authentication, triggers, and maintenance. **content_status:** `candidate` - Chapter: [English source](chapters/14-discover-and-audit-skills-EN.md) - Experiments: [Lab 004 — EN source](labs/lab-004-skill-selection-EN.md) · [Lab 005 — EN source](labs/lab-005-design-a-skill-EN.md) ### Chapter 15 — Research track: from question to auditable knowledge Narrow the research question; handle sources, citations, method, review, disclosure, and completeness. **content_status:** `candidate` - Chapter: [English source](chapters/15-research-track-EN.md) - Experiment: [Lab 008 — EN source](labs/lab-008-research-question-EN.md) ### Chapter 16 — Engineering track: from idea to reliable software Requirements, specifications, planning, incremental implementation, tests, debugging, review, release, and migration. **content_status:** `candidate` - Chapter: [English source](chapters/16-engineering-track-EN.md) - Experiment: [Lab 009 — EN source](labs/lab-009-engineering-lifecycle-EN.md) ### Chapter 17 — Marketing track: from product understanding to growth experiments Product context, audience, positioning, content, conversion, measurement, and attribution. **content_status:** `candidate` - Chapter: [English source](chapters/17-marketing-track-EN.md) - Experiment: [Lab 010 — EN source](labs/lab-010-product-context-EN.md) ### Chapter 18 — Content, design, data, and automation track Use the external ecosystem by task capability cluster instead of installing every Skill blindly. **content_status:** `candidate` - Chapter: [English source](chapters/18-content-design-data-automation-EN.md) - Experiment: [Lab 004 — EN source](labs/lab-004-skill-selection-EN.md) ## Part IV: From proficient use to organizational capability ### Chapter 19 — Evaluate models and workflows Build task sets, repeat experiments, score with human review, and classify errors. **content_status:** `candidate` - Chapter: [English source](chapters/19-evaluate-models-and-workflows-EN.md) - Experiments: [Lab 003 — EN source](labs/lab-003-evidence-review-EN.md) · [Lab 009 — EN source](labs/lab-009-engineering-lifecycle-EN.md) - Evaluation framework: [locale-neutral governance](../docs/quality/evaluation-framework.md) ### Chapter 20 — Build a personal Codex work system Project context, memory, templates, recurring workflows, and retrospectives. **content_status:** `candidate` - Chapter: [English source](chapters/20-personal-codex-work-system-EN.md) - Experiments: [Lab 001 — EN source](labs/lab-001-first-safe-task-EN.md) · [Lab 010 — EN source](labs/lab-010-product-context-EN.md) ### Chapter 21 — Build a team capability system Shared Skills, `AGENTS.md`, permissions, evaluation, review, contribution, and versioning. **content_status:** `candidate` - Chapter: [English source](chapters/21-team-capability-system-EN.md) - Experiment: [Lab 012 — EN source](labs/lab-012-team-capability-migration-EN.md) ### Chapter 22 — Continuous update and future-proofing Identify volatile facts, update sources, migrate models, audit tools, and remove obsolete capabilities. **content_status:** `candidate` · related volatile claim: `claim_status: current | disputed` - Chapter: [English source](chapters/22-continuous-update-and-future-proofing-EN.md) - Experiments: [Lab 008 — EN source](labs/lab-008-research-question-EN.md) · [Lab 010 — EN source](labs/lab-010-product-context-EN.md) ## Experiment index and status boundaries The repository has 18 real experiment files. Each remains `draft` and `run_status: not_run`; a directory link is an entry point, not evidence that the experiment or its learning outcome is verified. | Experiment | Focus | State | Entry | |---|---|---|---| | lab-001 | First safe task | `draft` · `not_run` | [EN source](labs/lab-001-first-safe-task-EN.md) | | lab-002 | Task protocol | `draft` · `not_run` | [EN source](labs/lab-002-task-protocol-EN.md) | | lab-003 | Evidence review | `draft` · `not_run` | [EN source](labs/lab-003-evidence-review-EN.md) | | lab-004 | Skill selection | `draft` · `not_run` | [EN source](labs/lab-004-skill-selection-EN.md) | | lab-005 | Skill design | `draft` · `not_run` | [EN source](labs/lab-005-design-a-skill-EN.md) | | lab-006 | Agent stop conditions | `draft` · `not_run` | [EN source](labs/lab-006-agent-stop-conditions-EN.md) | | lab-007 | Action boundaries | `draft` · `not_run` | [EN source](labs/lab-007-action-boundaries-EN.md) | | lab-008 | Research question | `draft` · `not_run` | [EN source](labs/lab-008-research-question-EN.md) | | lab-009 | Engineering lifecycle | `draft` · `not_run` | [EN source](labs/lab-009-engineering-lifecycle-EN.md) | | lab-010 | Product context | `draft` · `not_run` | [EN source](labs/lab-010-product-context-EN.md) | | lab-011 | GPT, Codex, tools, and Agents | `draft` · `not_run` | [Lab 011 — EN source](labs/lab-011-gpt-codex-boundaries-EN.md) | | lab-012 | Team capability migration | `draft` · `not_run` | [EN source](labs/lab-012-team-capability-migration-EN.md) | | lab-013 | Auditable L3 vertical slice | `draft` · `not_run` | [EN source](labs/lab-013-l3-vertical-slice-EN.md) | | lab-014 | Resume reconciliation | `draft` · `not_run` | [EN source](labs/lab-014-resume-reconciliation-EN.md) | | lab-015 | Evidence delivery | `draft` · `not_run` | [EN source](labs/lab-015-evidence-delivery-EN.md) | | lab-016 | Side-effect boundary | `draft` · `not_run` | [EN source](labs/lab-016-side-effect-boundary-EN.md) | | lab-017 | Skill discovery audit | `draft` · `not_run` | [EN source](labs/lab-017-skill-discovery-audit-EN.md) | | lab-018 | Language transfer under a fixed practice contract | `draft` · `not_run` | [EN source](labs/lab-018-language-transfer-EN.md) | ## Evaluation, status, and real-problem research - [Experiment index — English source](labs/README-EN.md): 18 real experiment files, levels, domains, migration focus, and `lab_status`. - [Content integration matrix — locale-neutral governance](../docs/content-matrix.md): capability mappings and the additional capability introduced when themes repeat. - [Evaluation framework — locale-neutral governance](../docs/quality/evaluation-framework.md): acceptance gates for content and capability. - [Learning path contract — locale-neutral governance](../docs/governance/learning-path.yaml): levels, primary experiments, supporting experiments, and progression gates. - [Codex real-user problem research — locale-neutral research](../docs/research/field-problems-codex.md): public problem entry points without claiming official root causes. - [Real-problem research index — locale-neutral research](../docs/research/field-problems-index-2026-08-10.md): maps FP, FP-S, FUP, forum findings, and chapter/lab locations. - [Forums and public-issue research — locale-neutral research](../docs/research/field-problems-forums-2026-08-10.md): reliable Stack Overflow API/pages and GitHub issue summaries. - [Official baseline research archive — locale-neutral research](../docs/research/openai-codex-baseline.md): source boundary for volatile claims. The independent primary experiments for L0, L3, and L6 are [Lab 011 — EN source](labs/lab-011-gpt-codex-boundaries-EN.md), [Lab 013 — EN source](labs/lab-013-l3-vertical-slice-EN.md), and [Lab 012 — EN source](labs/lab-012-team-capability-migration-EN.md).