sergstack · sergstack · Aug 28, 2026 · Aug 28, 2026
@@ -3,26 +3,26 @@
 /AGENTS.md @sergstack
 /CLAUDE.md @sergstack
 /GOAL_MODE.md @sergstack
-/GOAL_PACKS.md @sergstack
-/COMMAND_SURFACE.md @sergstack
-/CONTEXT_PACK_STANDARD.md @sergstack
+/docs/standards/GOAL_PACKS.md @sergstack
+/docs/standards/COMMAND_SURFACE.md @sergstack
+/docs/standards/CONTEXT_PACK_STANDARD.md @sergstack
 /MASTER_STATUS.md @sergstack
 /CURRENT_STATUS.md @sergstack
 /SYNC_CONTRACT.md @sergstack
 /README.md @sergstack
 /MANIFEST.md @sergstack
 /MANIFEST.json @sergstack
 /PROJECT_REGISTRY.md @sergstack
-/PARENT_CHILD_ISSUE_GATE_STANDARD.md @sergstack
-/EXISTING_SCRIPT_CONTROLLED_REFACTOR_STANDARD.md @sergstack
+/docs/standards/PARENT_CHILD_ISSUE_GATE_STANDARD.md @sergstack
+/docs/standards/EXISTING_SCRIPT_CONTROLLED_REFACTOR_STANDARD.md @sergstack
 /ARCHIVE_MAP.md @sergstack
 /UPLOAD_GUIDE.md @sergstack
 /REPO_PATHS.md @sergstack
 /AUTONOMOUS_EXECUTION_STANDARD.md @sergstack
 /AUTONOMOUS_EXECUTION_EXTENSION_CONTRACT.md @sergstack
 /ROUTING_RULES.md @sergstack
 /HANDOFF_STYLE_STANDARD.md @sergstack
-/PROMPT_QA_FACTORY.md @sergstack
+/docs/standards/PROMPT_QA_FACTORY.md @sergstack
 /PROJECT_CAPABILITIES.yaml @sergstack
 /knowledge_bundle_manifest.json @sergstack
 /docs/operations/CHATGPT_PROJECT_SYNC_CHECKLIST.md @sergstack

diff --git a/ChatGPT/[AI OS]/Knowledge/AIOS_01_ROUTING_AND_WORKFLOW_BUNDLE_SEMANTICS.md b/ChatGPT/[AI OS]/Knowledge/AIOS_01_ROUTING_AND_WORKFLOW_BUNDLE_SEMANTICS.md
@@ -9,7 +9,7 @@ Legacy bundle provenance: `ChatGPT/[AI OS]/Knowledge_Bundles/AIOS_01_ROUTING_AND
 If raw-input triage and AI OS scoped routing differ, use Inbox Router for
 triage and this file for AI OS evidence/governance scope.
 
-## Legacy section: `EXISTING_SCRIPT_CONTROLLED_REFACTOR_STANDARD.md`
+## Legacy section: `docs/standards/EXISTING_SCRIPT_CONTROLLED_REFACTOR_STANDARD.md`
 
 This is an engineering/Codex standard, not Analytics methodology. It does not define business logic, metrics, formulas, or analytical conclusions.
 Use `Existing Script Controlled Refactor Standard` when an existing script, CLI, notebook-exported script, or pipeline already runs or has known useful output, and Sergey wants cleanup, controlled refactor, modularization, simplification, or removal of obsolete internal code without behavior loss.
@@ -20,7 +20,7 @@ Any output contract change requires separate explicit acceptance.
 Use the smallest meaningful safety checks: existing tests, focused regression tests, golden-output comparison, smoke run/dry-run, schema check, row-count or reconciliation-total check, CLI help/entrypoint check, artifact validation, `git diff --check`, or repo-specific validation scripts.
 Allowed only after baseline, output contract, and safety checks exist: extract functions, split internal modules, rename internal helpers, isolate CLI/config/IO/transform/validate/report layers, remove truly dead or obsolete code, remove debug-only branches outside accepted behavior, replace duplicated internal logic with an equivalent helper, clarify comments/docstrings, and add focused tests around preserved behavior.
 Forbidden without separate explicit acceptance: behavior changes, output contract changes, schema/column/file-format/file-location changes, metric/formula/business-rule/financial-control changes, dependency additions, provider/API behavior changes, real provider/API execution, migrations, production/runtime/deploy changes, broad rewrite, deleting tests/QA/validation, source data mutation, runtime artifacts outside accepted fixture policy, autonomous loops, embeddings, semantic search, vector DB, or web UI.
-For large or risky refactors, use `Parent / Child Issue Gate Standard` from `PARENT_CHILD_ISSUE_GATE_STANDARD.md`. Do not duplicate the full parent/child standard.
+For large or risky refactors, use `Parent / Child Issue Gate Standard` from `docs/standards/PARENT_CHILD_ISSUE_GATE_STANDARD.md`. Do not duplicate the full parent/child standard.
 Typical child issues: baseline and output contract; safety tests or golden checks; dead-code cleanup; module extraction; final before/after comparison. Do not start downstream cleanup/refactor child issues until baseline and safety-test child issues are accepted or merged.
 Pass only when baseline behavior is captured, output contract is explicit, safety tests or comparison checks exist and run, cleanup stays in scope, before/after output is compared, output contract is preserved unless separately accepted, forbidden changes are absent, and final report lists changed files, checks, risks, rollback, and acceptance status.
 Stop when current behavior cannot be run/inspected/baselined, output contract cannot be inferred safely, no meaningful safety test/comparison path exists, required input data is missing and no safe fixture can be used, the task requires secrets/local absolute paths/production systems/real provider/API/source mutation, preserving behavior conflicts with requested cleanup, or the requested change would alter schema, metrics, formulas, business rules, APIs, file formats, column order, or output locations without separate acceptance.
diff --git a/...PT/[AI OS]/Knowledge/AIOS_04_GOAL_PACKS_AND_COMMAND_SURFACE_BUNDLE_SEMANTICS.md b/...PT/[AI OS]/Knowledge/AIOS_04_GOAL_PACKS_AND_COMMAND_SURFACE_BUNDLE_SEMANTICS.md
@@ -3,7 +3,7 @@
 Canonical source created during Issue #285 provenance migration.
 Legacy bundle provenance: `ChatGPT/[AI OS]/Knowledge_Bundles/AIOS_04_GOAL_PACKS_AND_COMMAND_SURFACE.md`.
 
-## Legacy section: `GOAL_PACKS.md`
+## Legacy section: `docs/standards/GOAL_PACKS.md`
 
 Goal Packs are reusable workflows for broad goals. Sergey starts from intent; AI-OS, LLM, Analytics, or Codex infer safe execution details.
 ### Active packs
@@ -29,21 +29,21 @@ Goal Packs are reusable workflows for broad goals. Sergey starts from intent; AI
 | `weekly_ai_os_review` | `[AI OS]` -> `[Thinking]` / `[Codex]` | review drift and choose one next useful improvement | do not create status ledgers or operating journals |
 | `reconciliation_builder` | `[Analytics]` -> `[Codex]` | build or improve reconciliation workflow | keep numeric logic deterministic and source layers explicit |
 
-## Legacy section: `COMMAND_SURFACE.md`
+## Legacy section: `docs/standards/COMMAND_SURFACE.md`
 
 One-touch command map for AI-OS, Stream Deck buttons, and quick prompts. Each command starts from the result Sergey wants, not from an atomic task form.
 | `Autoloop Analysis` | `[Analytics]` | question, data, QA criteria, stop conditions | supervised revise/rerun loop or blocker report | `autoloop` |
 Usage:
 
-## Legacy section: `CONTEXT_PACK_STANDARD.md`
+## Legacy section: `docs/standards/CONTEXT_PACK_STANDARD.md`
 
 Context Packs are compact inputs for AI-OS, LLM, Analytics, and Codex workflows. They contain the context needed for the next decision or output, not every available file.
 ### Minimal schema
 ### Guidance
 ### Quality gate
 A Context Pack is ready when the goal is clear, sources are named, facts and assumptions are separated, constraints and forbidden actions are visible, the expected output is specific, and the receiving project can act without asking Sergey to write an atomic task package.
 
-## Legacy section: `PROMPT_QA_FACTORY.md`
+## Legacy section: `docs/standards/PROMPT_QA_FACTORY.md`
 
 Core loop:
 Applies to StreamDeck prompts, ChatGPT Project prompts, Codex prompts, Judge/Revisor prompts, and Analytics memo prompts.

@@ -38,7 +38,7 @@
 | `JUDGE_CALIBRATION.md` | judge rules | Как AI-OS использует LLM-as-a-Judge, не считая judge output объективной истиной | Перед использованием или изменением judge workflows |
 | `LOOP_ACCEPTANCE_CHECKLIST.md` | loop QA | Checklist для решения, безопасен ли supervised loop к запуску | Перед запуском или acceptance supervised loop |
 | `SKILLS_HOOKS_MCP_DECISION_MATRIX.md` | tooling governance | Decision matrix для skills, hooks, MCP tools и sub-agents как workflow aids | Когда нужно решить, добавлять или использовать workflow tooling |
-| `EXISTING_SCRIPT_CONTROLLED_REFACTOR_STANDARD.md` | engineering / Codex standard | Controlled cleanup/refactor of existing working scripts without behavior loss | Когда рабочий скрипт надо почистить: сначала baseline, output contract, safety tests, затем refactor |
+| `docs/standards/EXISTING_SCRIPT_CONTROLLED_REFACTOR_STANDARD.md` | engineering / Codex standard | Controlled cleanup/refactor of existing working scripts without behavior loss | Когда рабочий скрипт надо почистить: сначала baseline, output contract, safety tests, затем refactor |
 
 ## Существующие KB-файлы, которые должны остаться основой
 

diff --git a/ChatGPT/[AI OS]/Knowledge_Bundles/AIOS_01_ROUTING_AND_WORKFLOW.md b/ChatGPT/[AI OS]/Knowledge_Bundles/AIOS_01_ROUTING_AND_WORKFLOW.md
@@ -11,7 +11,7 @@ Compact upload artifact for [AI OS] covering routing, workflow, and engineering/
 - `ChatGPT/[AI OS]/Knowledge/AI_OS_WORKFLOW.md`
 - `ChatGPT/[AI OS]/Knowledge/KB_USAGE_RULES.md`
 - `ChatGPT/[AI OS]/Knowledge/ANALYTICAL_MEMO_FACTORY_CODEX_APP_WORKFLOW.md`
-- `EXISTING_SCRIPT_CONTROLLED_REFACTOR_STANDARD.md`
+- `docs/standards/EXISTING_SCRIPT_CONTROLLED_REFACTOR_STANDARD.md`
 - `ChatGPT/[AI OS]/Knowledge/AIOS_01_ROUTING_AND_WORKFLOW_BUNDLE_SEMANTICS.md`
 
 ## Upload target
@@ -23,7 +23,7 @@ ChatGPT Project Sources / Knowledge for `[AI OS]`.
 - production_promotion: no, unless explicitly accepted elsewhere
 - bundle_type: generated compact upload artifact
 - source_of_truth: declared granular source files
-- source_fingerprint: sha256:485f07d1596d2ebff146ea71793e07ceb6ee02ef0b09dd55ef7adb28bfd15330
+- source_fingerprint: sha256:e672f866db5838b3fa92df97e41e32d0ed1b3446d2df3ca8381e395ee061bac2
 - generator: scripts/build_knowledge_bundles.py
 
 ---
@@ -67,7 +67,7 @@ ChatGPT Project Sources / Knowledge for `[AI OS]`.
 | `JUDGE_CALIBRATION.md` | judge rules | Как AI-OS использует LLM-as-a-Judge, не считая judge output объективной истиной | Перед использованием или изменением judge workflows |
 | `LOOP_ACCEPTANCE_CHECKLIST.md` | loop QA | Checklist для решения, безопасен ли supervised loop к запуску | Перед запуском или acceptance supervised loop |
 | `SKILLS_HOOKS_MCP_DECISION_MATRIX.md` | tooling governance | Decision matrix для skills, hooks, MCP tools и sub-agents как workflow aids | Когда нужно решить, добавлять или использовать workflow tooling |
-| `EXISTING_SCRIPT_CONTROLLED_REFACTOR_STANDARD.md` | engineering / Codex standard | Controlled cleanup/refactor of existing working scripts without behavior loss | Когда рабочий скрипт надо почистить: сначала baseline, output contract, safety tests, затем refactor |
+| `docs/standards/EXISTING_SCRIPT_CONTROLLED_REFACTOR_STANDARD.md` | engineering / Codex standard | Controlled cleanup/refactor of existing working scripts without behavior loss | Когда рабочий скрипт надо почистить: сначала baseline, output contract, safety tests, затем refactor |
 ## Существующие KB-файлы, которые должны остаться основой
 Минимальный ожидаемый набор уже загруженной KB:
 ```text
@@ -397,7 +397,7 @@ Do not force the user into a manual loop where `[Analytics]` asks for Python out
 - production_promotion: no
 - source_of_truth: this file plus the granular Analytics and Codex workflow files
 
-## From: `EXISTING_SCRIPT_CONTROLLED_REFACTOR_STANDARD.md`
+## From: `docs/standards/EXISTING_SCRIPT_CONTROLLED_REFACTOR_STANDARD.md`
 
 # Existing Script Controlled Refactor Standard
 ## Purpose
@@ -515,7 +515,7 @@ cli / entrypoint
 ```
 This split is recommended, not mandatory. Use the smallest structure that makes behavior safer and clearer.
 ## Parent/child decomposition for large risky refactors
-For large or risky refactors, use `Parent / Child Issue Gate Standard` from `PARENT_CHILD_ISSUE_GATE_STANDARD.md`.
+For large or risky refactors, use `Parent / Child Issue Gate Standard` from `docs/standards/PARENT_CHILD_ISSUE_GATE_STANDARD.md`.
 Do not duplicate the full parent/child standard here.
 A large refactor should usually be split into child issues such as:
 1. baseline and output contract;
@@ -574,7 +574,7 @@ Legacy bundle provenance: `ChatGPT/[AI OS]/Knowledge_Bundles/AIOS_01_ROUTING_AND
 `ChatGPT/[Inbox Router]/Knowledge/ROUTING_RULES.md`.
 If raw-input triage and AI OS scoped routing differ, use Inbox Router for
 triage and this file for AI OS evidence/governance scope.
-## Legacy section: `EXISTING_SCRIPT_CONTROLLED_REFACTOR_STANDARD.md`
+## Legacy section: `docs/standards/EXISTING_SCRIPT_CONTROLLED_REFACTOR_STANDARD.md`
 This is an engineering/Codex standard, not Analytics methodology. It does not define business logic, metrics, formulas, or analytical conclusions.
 Use `Existing Script Controlled Refactor Standard` when an existing script, CLI, notebook-exported script, or pipeline already runs or has known useful output, and Sergey wants cleanup, controlled refactor, modularization, simplification, or removal of obsolete internal code without behavior loss.
 Do not use this standard for greenfield implementation, broken-script recovery, requested behavior/schema/metric/formula/API/provider/output-contract changes, production deploy, migration, source mutation, Safe Apply, real provider/API execution without approval, or cases where current behavior cannot be baselined.
@@ -584,7 +584,7 @@ Any output contract change requires separate explicit acceptance.
 Use the smallest meaningful safety checks: existing tests, focused regression tests, golden-output comparison, smoke run/dry-run, schema check, row-count or reconciliation-total check, CLI help/entrypoint check, artifact validation, `git diff --check`, or repo-specific validation scripts.
 Allowed only after baseline, output contract, and safety checks exist: extract functions, split internal modules, rename internal helpers, isolate CLI/config/IO/transform/validate/report layers, remove truly dead or obsolete code, remove debug-only branches outside accepted behavior, replace duplicated internal logic with an equivalent helper, clarify comments/docstrings, and add focused tests around preserved behavior.
 Forbidden without separate explicit acceptance: behavior changes, output contract changes, schema/column/file-format/file-location changes, metric/formula/business-rule/financial-control changes, dependency additions, provider/API behavior changes, real provider/API execution, migrations, production/runtime/deploy changes, broad rewrite, deleting tests/QA/validation, source data mutation, runtime artifacts outside accepted fixture policy, autonomous loops, embeddings, semantic search, vector DB, or web UI.
-For large or risky refactors, use `Parent / Child Issue Gate Standard` from `PARENT_CHILD_ISSUE_GATE_STANDARD.md`. Do not duplicate the full parent/child standard.
+For large or risky refactors, use `Parent / Child Issue Gate Standard` from `docs/standards/PARENT_CHILD_ISSUE_GATE_STANDARD.md`. Do not duplicate the full parent/child standard.
 Typical child issues: baseline and output contract; safety tests or golden checks; dead-code cleanup; module extraction; final before/after comparison. Do not start downstream cleanup/refactor child issues until baseline and safety-test child issues are accepted or merged.
 Pass only when baseline behavior is captured, output contract is explicit, safety tests or comparison checks exist and run, cleanup stays in scope, before/after output is compared, output contract is preserved unless separately accepted, forbidden changes are absent, and final report lists changed files, checks, risks, rollback, and acceptance status.
 Stop when current behavior cannot be run/inspected/baselined, output contract cannot be inferred safely, no meaningful safety test/comparison path exists, required input data is missing and no safe fixture can be used, the task requires secrets/local absolute paths/production systems/real provider/API/source mutation, preserving behavior conflicts with requested cleanup, or the requested change would alter schema, metrics, formulas, business rules, APIs, file formats, column order, or output locations without separate acceptance.