ytvee · Oct 28, 2026 · Oct 28, 2026 · Oct 28, 2026 · Oct 28, 2026 · Oct 28, 2026
diff --git a/.claude-plugin/plugin.json b/.claude-plugin/plugin.json
@@ -1,6 +1,6 @@
 {
   "name": "webdev-agent-kit",
-  "version": "1.8.0",
+  "version": "1.9.0",
   "description": "Portable frontend workflow skills for planning, implementation, review, and visual QA.",
   "author": {
     "name": "Yevgeniy Tyan"

diff --git a/.codex-plugin/plugin.json b/.codex-plugin/plugin.json
@@ -1,6 +1,6 @@
 {
   "name": "webdev-agent-kit",
-  "version": "1.8.0",
+  "version": "1.9.0",
   "description": "Frontend workflow skills for React and Next.js planning, implementation, review, and visual QA.",
   "skills": "./skills/"
 }
diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
@@ -138,10 +138,11 @@ jobs:
           test -d dist/claude
           test -d dist/claude-code
           test -d dist/cursor
+          test -d dist/kimi
           test -d dist/vs-code-codex
           test -d dist/vs-code-claude
 
-          for target in codex claude claude-code cursor vs-code-codex vs-code-claude; do
+          for target in codex claude claude-code cursor kimi vs-code-codex vs-code-claude; do
             test -f "dist/${target}/LICENSE"
             test -f "dist/${target}/tool-capabilities-manifest.json"
             test ! -e "dist/${target}/project"
@@ -210,11 +211,13 @@ jobs:
             "release-artifacts/webdev-agent-kit-codex-${RELEASE_TAG}.tar.gz"
             "release-artifacts/webdev-agent-kit-claude-code-${RELEASE_TAG}.tar.gz"
             "release-artifacts/webdev-agent-kit-cursor-${RELEASE_TAG}.tar.gz"
+            "release-artifacts/webdev-agent-kit-kimi-${RELEASE_TAG}.tar.gz"
             "release-artifacts/webdev-agent-kit-vs-code-codex-${RELEASE_TAG}.tar.gz"
             "release-artifacts/webdev-agent-kit-vs-code-claude-${RELEASE_TAG}.tar.gz"
             "release-artifacts/webdev-agent-kit-codex.tar.gz"
             "release-artifacts/webdev-agent-kit-claude-code.tar.gz"
             "release-artifacts/webdev-agent-kit-cursor.tar.gz"
+            "release-artifacts/webdev-agent-kit-kimi.tar.gz"
             "release-artifacts/webdev-agent-kit-vs-code-codex.tar.gz"
             "release-artifacts/webdev-agent-kit-vs-code-claude.tar.gz"
             "release-artifacts/SHA256SUMS"

diff --git a/AGENTS.md b/AGENTS.md
@@ -11,6 +11,7 @@ tags:
     - 'docs/entrypoint'
 parent: []
 related:
+    - '[[adapters/kimi|Kimi Code Adapter]]'
     - '[[common/webdev-role-selection]]'
     - '[[common/codex-model-routing-policy]]'
     - '[[skills/frontend-prototype-explorer/SKILL|Frontend Prototype Explorer]]'
@@ -62,7 +63,9 @@ Load in this order, stopping when the next safe action is clear:
 4. Verified project instructions and only the affected `project/**` facts.
 5. The smallest matching `skills/**/SKILL.md` and only references it requires for the active slice.
 
-Do not load every skill, reference, rule, overlay, README or target for routing. Consult `common/prompt-intent-routing-rules.md` only for unclear workflow weight. `tool-capabilities-manifest.json` defines needs; callable native tools can satisfy them. Names and configuration do not prove availability.
+Load narrowly. Use `common/prompt-intent-routing-rules.md` for unclear intent.
+`tool-capabilities-manifest.json` defines needs; callable tools, not names or
+configuration, establish availability.
 
 ## Skill Discovery
 
@@ -77,7 +80,7 @@ Select skills by name and description; explicit invocation is optional. Each `SK
 - Build: `frontend-architecture-planner` defines ownership; `greenfield-project-builder` plans a first vertical slice; `frontend-layout-implementer` implements an approved spec in its target stack.
 - Quality: `frontend-bugfix-debugger` fixes defects; `frontend-refactor-surgeon` preserves behavior; `frontend-linter-manager` runs or repairs scoped lint; `frontend-visual-qa` checks rendering; `frontend-quality-reviewer` reviews independently.
 - Context and tooling: `project-onboarding-adapter` initializes pointers and local facts; `project-context-adapter` refreshes them; `mcp-toolchain-manager` maps tool capabilities; `pattern-library-manager` maintains reusable patterns; `agent-rules-skill-author` maintains this bundle.
-- Learning: `agent-self-improvement` (opt-in).
+- Learning: `agent-self-improvement` saves explicitly requested experience to the learning database, including natural-language requests. Never substitute project-rule edits.
 - Feedback: `webdev-feedback` files requested issues. Updates: `webdev-kit-updater` preserves local changes.
 
 ## Change Boundaries
@@ -88,6 +91,10 @@ Use targeted source or platform documentation only when current facts affect the
 
 ## Verification And Output
 
-Run the smallest relevant existing check for the changed surface. Bound retries, change strategy after a failed approach, and never claim a check, tool, screenshot, or result without evidence. Report exact blocked checks when no honest fallback exists.
+Omit routine model/version/effort announcements; include only on request or for
+model diagnostics. Preserve verified trace/evaluation metadata.
+
+Run relevant existing checks, bound retries, change failed approaches and report
+blocked checks. Never claim tools, screenshots or results without evidence.
 
 Apply `common/token-economy-rules.md`. Return only facts that affect understanding, confidence, or the next action. For changed work, default to `Changed`, `Verified`, and `Blocked` when applicable; lightweight final answers stay within 180 words. Omit request echo, skill names, workflow narration, empty sections, raw logs, and irrelevant skipped-work notes; never hide errors, unknowns, blocked verification, or approval needs.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -11,6 +11,7 @@ tags:
     - 'release-management'
 parent: []
 related:
+    - '[[docs/release/1.9.0-plan]]'
     - '[[docs/install/project-memory]]'
     - '[[docs/release/1.8.0-plan]]'
     - '[[docs/install/model-capability-guide]]'
@@ -38,6 +39,37 @@ Use this changelog for source-bundle and distribution-target changes that affect
 
 ### Added
 
+- Consolidated STEM learning workflows on the current Kit: scoped evidence,
+  diagnosis, risk-based evaluation, processing receipts, recovery and snapshots.
+  Full transferable-source inventory documents preservation and approved changes.
+- Optional Langfuse adaptation with credential reuse or an opened private key file,
+  explicit connection/decline, supported hooks, stable project tags and probe readback.
+- Natural-language memory writes with local storage, readback and no implicit
+  rule edits, analysis, capture, tracing or synchronization.
+- Kimi Code distribution, installation guide and archive validation.
+  Release publication includes both versioned and stable-name Kimi archives.
+
+### Changed
+
+- New project memory defaults to local storage with capture, telemetry and exchange
+  off and no preset remote. Existing choices and legacy history are preserved.
+- Project CLI/hook storage resolution is consistent across both project flag forms,
+  working directories and explicit overrides; migration remains project-scoped.
+- Triage skips exactly curated source events; explicit reconsideration is separate.
+  Long sessions use bounded stable pages instead of unbounded context exports.
+- Snapshot exchange uses the verified SQLite repository, rejects conflicting legacy
+  addresses and requires separate completion-stage consent. No background learning.
+- Learning telemetry queries selected event windows and reuses bounded cached data;
+  missing evidence falls back to the database without weakening acceptance criteria.
+- Routine model/version/effort boilerplate is omitted from user-facing responses.
+  Runtime/evaluation metadata stays intact. Local memory is protected during updates.
+- Source, behavior and real-service verification are reported separately; tests do
+  not claim model improvement or live client/account activation.
+
+## 1.8.0 — 2026-10-23
+
+### Added
+
 - Memory quality audit, evidence checks and supported acceptance bases for new
   verified lessons. Unsupported historical lessons stay stored and are excluded
   from ordinary recall.

diff --git a/adapters/kimi.md b/adapters/kimi.md
@@ -0,0 +1,43 @@
+---
+id: 'agents.adapters.kimi'
+title: 'Kimi Code Adapter'
+doc_type: 'adapter'
+layer: 'adapter'
+status: 'active'
+publishable: true
+local_only: false
+tags:
+    - 'agents/adapter'
+parent:
+    - '[[AGENTS|Canonical Agent Policy]]'
+related: []
+depends_on:
+    - '[[common/core/runtime-core-policy|Portable Runtime Core]]'
+---
+
+# Kimi Code Adapter
+
+Use the portable runtime core and host instructions. This adapter targets Kimi
+Code CLI with project `.agents/skills` discovery, including K3 when available
+in the user's account. Preserve selected models, permissions and authentication.
+
+Extract the Kimi project archive into the project root, preserving existing
+`.agents/project` and `.agents/learning` state. Start a fresh session and invoke
+`/skill:project-context-adapter` to verify skill discovery, references and tools.
+Each generated skill loads the policy through `${KIMI_SKILL_DIR}`. Do not replace
+an existing host AGENTS.md to activate this package.
+
+Use actually available filesystem, shell, browser and MCP capabilities. Subagents require verified availability and scoped handoffs. Do not reuse
+Codex/Claude configurations or install dependencies implicitly.
+
+Learning remains opt-in. Use portable Python memory commands and explicit canonical JSONL imports. This
+release does not install a Kimi capture hook or parse undocumented transcripts.
+Codex/OpenCode proposal runners are not evidence of a Kimi model evaluation;
+report an unavailable evaluator rather than claiming a live Kimi pass.
+
+## Official References
+
+- [Skills and project discovery](https://www.kimi.com/code/docs/en/kimi-code-cli/customization/skills.html)
+- [Agents and tool boundaries](https://www.kimi.com/code/docs/en/kimi-code-cli/customization/agents.html)
+
+Inspected 2026-10-28. Packaging checks do not prove live client behavior.
diff --git a/bundle-manifest.json b/bundle-manifest.json
@@ -1,6 +1,6 @@
 {
   "name": "webdev-agent-kit",
-  "version": "1.8.0",
+  "version": "1.9.0",
   "license": "Apache-2.0",
   "description": "Internal source and distribution inventory for the WebDev Agent Kit skill bundle.",
   "portable_core": "common/core/runtime-core-policy.md",
@@ -11,7 +11,8 @@
   "adapters": {
     "claude-code": "adapters/claude-code.md",
     "codex": "adapters/codex.md",
-    "cursor": "adapters/cursor.md"
+    "cursor": "adapters/cursor.md",
+    "kimi": "adapters/kimi.md"
   },
   "skills": [
     "agent-rules-skill-author",
@@ -62,6 +63,14 @@
       "entrypoint": ".cursor/rules/webdev-agent-kit.mdc",
       "skills_root": ".agents/skills",
       "validator": "scripts/validate_cursor_target.py"
+    },
+    "kimi": {
+      "kind": "project-bundle",
+      "output": "dist/kimi",
+      "install_mode": "project-root-extract",
+      "entrypoint": ".agents/AGENTS.md",
+      "skills_root": ".agents/skills",
+      "validator": "scripts/validate_kimi_target.py"
     }
   },
   "target_aliases": {

diff --git a/common/prompt-intent-routing-rules.md b/common/prompt-intent-routing-rules.md
@@ -32,6 +32,15 @@ Start with the cheapest path that can answer the user. Escalate only when target
 
 A narrow lookup must stay fast. A narrow bugfix must stay lightweight. A large ambiguous request must be planned before implementation.
 
+## Explicit Memory Requests
+
+Explicit requests to remember an incident, save feedback for learning or "запиши
+себе в базу" route to `agent-self-improvement` and its Manual Memory Writes flow.
+Use current-project SQLite memory, not project rules or a skill rewrite. Ask only
+for missing content when the available conversation cannot identify the lesson.
+Bare "remember" without a clear learning/experience context is not sufficient;
+an explicit request to edit project rules remains a rule-editing task.
+
 ## Question-First Gate
 
 When the user asks for options, tradeoffs, or approach before implementation, answer the question before editing files.

diff --git a/common/token-economy-rules.md b/common/token-economy-rules.md
@@ -73,6 +73,10 @@ Do not reduce thinking by skipping root-cause analysis, accessibility checks, se
 
 ## Output Economy
 
+Omit routine model/version/reasoning-effort announcements and unavailable-runtime
+boilerplate. Include these only when directly requested or material to diagnosing
+model configuration. This does not remove metadata from traces or evaluations.
+
 Default final reports are compact, fact-only, and evidence-first. Return only facts that affect the user's understanding, confidence, or next action.
 
 For tasks that changed or validated something, use this default shape:

diff --git a/docs/install/README.md b/docs/install/README.md
@@ -11,6 +11,7 @@ tags:
 parent:
     - '[[docs/README|WebDev Agent Kit Documentation]]'
 related:
+    - '[[docs/install/kimi|Kimi Code Installation]]'
     - '[[docs/install/opencode]]'
     - '[[docs/install/model-capability-guide]]'
     - '[[docs/install/roles]]'
@@ -25,7 +26,7 @@ depends_on: []
 
 # Installation Guides
 
-[← All documentation](../README.md)
+[в†ђ All documentation](../README.md)
 
 Choose the client you use to open your project:
 
@@ -51,8 +52,10 @@ is already installed, use [upgrade and rollback](upgrade.md).
 
 ## Video guides
 
-- [Cursor — text and video guide](cursor.md)
-- [VS Code Codex — text and video guide](vscode-codex.md)
+- [Cursor вЂ” text and video guide](cursor.md)
+- [VS Code Codex вЂ” text and video guide](vscode-codex.md)
 
 These pages are user guides, not runtime policy. Maintainers define package
 contents in `bundle-manifest.json` and check them with the archive validators.
+
+- [Kimi Code installation](kimi.md)
diff --git a/docs/install/first-run.md b/docs/install/first-run.md
@@ -26,6 +26,7 @@ download, file layout and first session.
 | --- | --- | --- | --- |
 | `opencode` (installer) | `codex` portable skills | host `.agents/` and `.opencode/plugins/` | `opencode.json` instructions + native plugin |
 | `codex` | `codex` | host `.agents/` | host `AGENTS.md` pointer to `.agents/AGENTS.md` |
+| `kimi` | `kimi` | Project archive | `.agents/skills` |
 | `vs-code-codex` | `codex` | host `.agents/` | same Codex contract |
 | `cursor` | `cursor` | host `.agents/` and `.cursor/` | `.cursor/rules/webdev-agent-kit.mdc` |
 | `claude-code` | `claude-code` | external `webdev-agent-kit/` | native plugin and local marketplace |
@@ -78,7 +79,7 @@ and its use is approved; never create a pointer to a missing `.agents/AGENTS.md`
 - Exercised: a small task has an observed result and retained evidence.
 
 Check discovery in a fresh session, then ask a read-only question such as
-“Which existing command checks types, and what source proves that?”.
+вЂњWhich existing command checks types, and what source proves that?вЂќ.
 Installation and discovery alone do not establish agent quality or cross-client
 compatibility. Do not run an app merely to claim onboarding success.
 

diff --git a/docs/install/kimi.md b/docs/install/kimi.md
@@ -0,0 +1,49 @@
+---
+id: 'agents.docs.install.kimi'
+title: 'Install WebDev Agent Kit for Kimi Code'
+doc_type: 'user-guide'
+layer: 'docs'
+status: 'active'
+publishable: true
+local_only: false
+tags:
+    - 'docs/install'
+    - 'client/kimi'
+parent:
+    - '[[docs/install/README|Installation Guides]]'
+related:
+    - '[[docs/install/gpt-model-routing]]'
+    - '[[docs/install/first-run|First Run And Installation Checks]]'
+    - '[[docs/install/upgrade|Upgrade And Rollback]]'
+    - '[[common/client-adaptation-policy|Client Adaptation Policy]]'
+    - '[[common/mcp-installation-policy|MCP Installation Policy]]'
+    - '[[docs/mcp/README|MCP for WebDev Agent Kit]]'
+depends_on: []
+---
+
+# Install WebDev Agent Kit for Kimi Code
+
+Complete the [first-run checks](first-run.md). Preserve existing installations
+using [upgrade and rollback](upgrade.md).
+
+1. Download the [Kimi project package](https://github.com/ytvee-dev/webdev-agent-kit/releases/latest/download/webdev-agent-kit-kimi.tar.gz).
+2. Extract it into the project root: skills belong under `.agents/skills`, with
+   policy, references and scripts alongside them inside `.agents/`.
+3. Preserve existing host instructions and `.agents/project/` and
+   `.agents/learning/` state. Keep private state excluded from Git.
+4. Start a fresh Kimi Code session in that project and invoke
+   `/skill:project-context-adapter`. Ask it to inspect the project and adapt
+   project facts, using `.agents/adapters/kimi.md`. Verify that the skill loads
+   its policy and supporting files. Do not overwrite an existing AGENTS.md.
+5. Keep the selected model and permissions. K3 requires availability in the
+   user's Kimi account; this package does not provision credentials or models.
+
+The adapter uses documented project skill discovery and `${KIMI_SKILL_DIR}`.
+Packaging checks do not prove live model behavior. Automatic Kimi capture and
+Kimi-specific learning evaluation are not included; use explicit canonical JSONL
+import and the portable memory CLI when requested. Existing Codex/OpenCode
+runners must not be reported as Kimi evaluation.
+
+[Official skill documentation](https://www.kimi.com/code/docs/en/kimi-code-cli/customization/skills.html)
+
+[All installation guides](README.md)
diff --git a/docs/install/project-memory.md b/docs/install/project-memory.md
@@ -21,6 +21,30 @@ The Kit remembers project experience by saving records and retrieving relevant
 ones for later tasks. It does not train model weights. A correction can become a
 candidate lesson; a useful lesson can later become a checked project rule.
 
+## Project-local defaults and explicit choices
+
+New state lives at `<project>/.agents/learning`; capture, telemetry and exchange
+start off. `--state PATH` takes precedence over `AGENT_LEARNING_STATE`, then the
+selected project. Both `--project ROOT` and `--project=ROOT` work from any directory.
+If the old shared store contains this project's history, migrate explicitly with
+`learning_project.py --project ROOT --legacy STATE`; the source is preserved.
+Status prints the actual state/database path. Kit updates preserve this directory.
+
+Ask the agent to save a correction to its learning database. It records structured
+experience and reports IDs; it does not rewrite project rules or start analysis.
+For long captured sessions, follow `context --limit 200` and its `next_cursor` via
+`--cursor TOKEN`. `run` skips curated source IDs; `--reconsider` requests reanalysis.
+
+During adaptation, choose Langfuse connection or decline. The agent offers saved
+keys or opens the private credential file, waits, sets project tags and checks
+readback. Disable tracing for one project with an explicit agent request; other
+projects and local memory remain available. A stored choice survives updates.
+
+Verify a private sync destination with `learning_memory.py configure --project ROOT
+--repository OWNER/REPO`. Only add `--sync-on-finish yes` when completion-stage
+upload is wanted; `no` disables it. Address configuration alone never uploads.
+Conflicting legacy addresses must be resolved explicitly before snapshot exchange.
+
 ## Upgrade existing memory
 
 Upgrade the Kit on every computer that exchanges the same project memory before