savant-code 0.0.19 → 0.0.20

package/.agents/skills/release-workflow/SKILL.md

@@ -1,69 +0,0 @@
----
-name: release-workflow
-description: Release cycle conventions including CHANGELOG format, README maintenance, version bumping, and git workflow.
----
-
-# Release Workflow Standards
-
-<!-- Load this for all projects — defines the mandatory release cycle -->
-
-## Mandatory Release Cycle
-
-Every time code is pushed:
-
-1. **Update CHANGELOG.md** — add entries for all changes since last push
-2. **Review README.md** — update test counts, version references, any stale data
-3. **Commit docs** — `git add CHANGELOG.md README.md && git commit -m "docs: ..."`
-4. **Push** — `git push`
-5. **Create/update release** — tag + release notes on GitHub (or equivalent)
-
-**Never push code without updating CHANGELOG + README first. Never skip the release.**
-
-## CHANGELOG Format
-
-Follow [Keep a Changelog](https://keepachangelog.com/en/1.1.0/):
-
-```markdown
-## [VERSION] — YYYY-MM-DD
-
-### Added
-- New features
-
-### Fixed
-- Bug fixes
-
-### Changed
-- Modifications to existing behavior
-
-### Removed
-- Deprecated features
-```
-
-Group entries by FID when applicable. Include file references in parentheses.
-
-## README Maintenance
-
-On every release, verify:
-
-- [ ] Version number matches Cargo.toml / package.json
-- [ ] Test count matches `cargo test` output
-- [ ] FID count matches `dev/fids/` directory
-- [ ] All referenced features still exist
-- [ ] No stale links or descriptions
-
-## AGENTS.md
-
-Each project should have an `AGENTS.md` at the root with:
-
-- Project-specific release workflow (may override this standard)
-- Build & test commands
-- Protocol version reference
-- Any project-specific conventions
-
-## Version Bumping
-
-- **Patch** (0.1.0 → 0.1.1): Bug fixes, small improvements
-- **Minor** (0.1.0 → 0.2.0): New features, non-breaking changes
-- **Major** (0.1.0 → 1.0.0): Breaking changes, architecture shifts
-
-Update `VERSION` file first, then propagate to all config files.

package/.bun-version

@@ -1 +0,0 @@
-1.3.14

package/.commandcode/settings.json

@@ -1,11 +0,0 @@
-{
-  "permissions": {
-    "allow": [
-      "Shell(npm:*)",
-      "Shell(git:*)",
-      "powershell"
-    ],
-    "deny": [],
-    "defaultMode": "default"
-  }
-}

package/.commandcode/taste/workflow/taste.md

@@ -1,6 +0,0 @@
-- Prefers reading the specified project guidance file in full from start to finish before beginning the investigation. Confidence: 0.9
-- When debugging an npm-installed product, expects the investigation to distinguish the published package and downloaded runtime artifact from the local source checkout, rather than attributing issues to local code without evidence. Confidence: 0.9
-- Under an imminent launch deadline, prioritizes the smallest launch-safe fix and focused release smoke checks over broader changes. Confidence: 0.85
-- Requires hotfix releases to use a new patch version rather than reusing the existing release; in this case, the hotfix target is explicitly `0.0.18`. Confidence: 0.95
-- For release hotfixes, expects a comprehensive update across README and all relevant documentation, not only code and version metadata. Confidence: 0.95
-- Uses automation level 3 when asking to proceed, indicating a preference for the agent to carry out the authorized release workflow autonomously rather than stopping for routine confirmations. Confidence: 0.85

package/.env.example

@@ -1,95 +0,0 @@
-# =============================================================================
-# Savant Code CLI — Local Development Environment Template
-# =============================================================================
-# COPY NOTICE:
-# Copy this file to `.env.local` before local development:
-#
-#   cp .env.example .env.local       # macOS/Linux
-#   copy .env.example .env.local     # Windows Command Prompt
-#
-# `.env.local` is gitignored. Replace dummy credential values with real local
-# values only; never commit `.env.local` or paste live credentials into source.
-# This file intentionally preserves the complete configuration surface used by
-# the local CLI, direct-provider mode, optional gateway providers, backend mode,
-# release-adjacent local configuration, and required client environment validation.
-#
-# All credential values below are safe placeholders. The variable names and
-# commented state below mirror the local development configuration; do not
-# rename or remove entries when creating `.env.local`.
-
-# -----------------------------------------------------------------------------
-# Core environment
-# -----------------------------------------------------------------------------
-NEXT_PUBLIC_CB_ENVIRONMENT=dev
-NEXT_PUBLIC_WEB_PORT=3000
-
-# -----------------------------------------------------------------------------
-# App URLs
-# -----------------------------------------------------------------------------
-# Used by getWebsiteUrl() for non-inference backend calls. With direct
-# OpenRouter inference enabled, remaining backend calls still use this URL.
-NEXT_PUBLIC_SAVANT_CODE_APP_URL=http://localhost:3000
-
-# Optional separate free product URL.
-# NEXT_PUBLIC_FREEBUFF_APP_URL=http://localhost:3001
-
-# -----------------------------------------------------------------------------
-# Analytics / observability (dummy values)
-# -----------------------------------------------------------------------------
-NEXT_PUBLIC_POSTHOG_API_KEY=phc_dummy_replace_me
-NEXT_PUBLIC_POSTHOG_HOST_URL=http://localhost:4000
-NEXT_PUBLIC_GRAVITY_PIXEL_ID=00000000-0000-0000-0000-000000000000
-
-# -----------------------------------------------------------------------------
-# Support / billing (dummy values)
-# -----------------------------------------------------------------------------
-NEXT_PUBLIC_SUPPORT_EMAIL=replace-me@example.com
-NEXT_PUBLIC_STRIPE_PUBLISHABLE_KEY=pk_dummy_replace_me
-NEXT_PUBLIC_STRIPE_CUSTOMER_PORTAL=http://localhost:3000/portal
-NEXT_PUBLIC_GOOGLE_SITE_VERIFICATION_ID=dummy_replace_me
-
-# -----------------------------------------------------------------------------
-# Inference — OpenRouter direct mode
-# -----------------------------------------------------------------------------
-# Set DIRECT_PROVIDER and INFERENCE_BASE_URL to bypass the Savant Code backend
-# for inference. The SDK resolves the OpenRouter credential in this order:
-#   1. OR_MASTER_KEY — exchanges for a regular key via /api/v1/keys
-#   2. OPENROUTER_API_KEY — regular OpenRouter key
-#   3. INFERENCE_API_KEY — SDK-specific inference key
-DIRECT_PROVIDER=openrouter
-INFERENCE_BASE_URL=https://openrouter.ai/api/v1
-OR_MASTER_KEY=sk-or-master-dummy-replace-me
-
-# Leave INFERENCE_BASE_URL unset to use the Savant Code backend by default.
-
-# -----------------------------------------------------------------------------
-# Inference — Gateway providers (optional)
-# -----------------------------------------------------------------------------
-# TokenRouter: gateway for Kimi, DeepSeek V4, Qwen, GLM, MiniMax, GPT-OSS
-# https://tokenrouter.me/v1
-TOKENROUTER_API_KEY=sk-tokenrouter-dummy-replace-me
-
-# NVIDIA NIM: gateway for GLM-5.2, Llama, Nemotron, DeepSeek V3, Qwen, etc.
-# https://integrate.api.nvidia.com/v1
-NVIDIA_API_KEY=nvapi-dummy-replace-me
-
-# OpenCode Go: subscription gateway for curated open models (dual-protocol)
-# https://opencode.ai/zen/go/v1/
-OPENCODE_GO_API_KEY=sk-opencode-go-dummy-replace-me
-
-# CommandCode: OpenAI-compatible hosted inference
-COMMAND_CODE_API_KEY=commandcode-dummy-replace-me
-
-# -----------------------------------------------------------------------------
-# Auth — Savant Code backend mode
-# -----------------------------------------------------------------------------
-# SAVANT_CODE_API_KEY is used for Savant Code backend calls. When
-# INFERENCE_BASE_URL is set, local direct-provider development can bypass this
-# key. Uncomment and replace the dummy value when using backend mode.
-# SAVANT_CODE_API_KEY=
-
-# -----------------------------------------------------------------------------
-# Worker / deployment integration (dummy value)
-# -----------------------------------------------------------------------------
-# Worker/deployment credential used by local deployment integrations.
-AMAZON_WORKER=amazon-worker-dummy-replace-me

package/.gitattributes

@@ -1,43 +0,0 @@
-# Auto-detect text files and normalize line endings to LF
-* text=auto eol=lf
-
-# Explicitly declare text files
-*.ts text eol=lf
-*.tsx text eol=lf
-*.js text eol=lf
-*.jsx text eol=lf
-*.json text eol=lf
-*.md text eol=lf
-*.yaml text eol=lf
-*.yml text eol=lf
-*.toml text eol=lf
-*.css text eol=lf
-*.html text eol=lf
-*.sh text eol=lf
-*.bash text eol=lf
-*.py text eol=lf
-*.go text eol=lf
-*.rs text eol=lf
-*.sql text eol=lf
-*.env text eol=lf
-*.gitignore text eol=lf
-*.gitattributes text eol=lf
-
-# Denote files that are binary and should not be modified
-*.png binary
-*.jpg binary
-*.jpeg binary
-*.gif binary
-*.ico binary
-*.svg text eol=lf
-*.woff binary
-*.woff2 binary
-*.ttf binary
-*.eot binary
-*.wasm binary
-
-# Lock files — do not normalize (upstream format)
-bun.lock -text
-package-lock.json -text
-yarn.lock -text
-pnpm-lock.yaml -text

package/.githooks/pre-push

@@ -1,35 +0,0 @@
-#!/usr/bin/env bash
-# Pre-push hook — enforce the hard lint gates before anything reaches the remote.
-#
-# Runs the same gates as `protocol.config.yaml` (lint + lint_md):
-#   - bun x eslint . --max-warnings 0   (zero-tolerance ESLint, FID-069)
-#   - bun run lint:md                   (markdownlint on governed docs)
-#
-# Wired automatically by the root `prepare` script (`git config core.hooksPath
-# .githooks`), which runs on every `bun install`. Skip with `--no-verify`.
-
-set -uo pipefail
-
-if ! command -v bun >/dev/null 2>&1; then
-  echo "⚠️  pre-push: bun not found — skipping lint gates (run: bun install)"
-  exit 0
-fi
-
-echo "🔍 pre-push: running hard lint gates (eslint --max-warnings 0 + markdownlint)..."
-
-bun x eslint . --max-warnings 0
-eslint_status=$?
-if [ "$eslint_status" -ne 0 ]; then
-  echo "❌ pre-push: ESLint reported errors/warnings — fix before pushing (or use --no-verify)" >&2
-  exit "$eslint_status"
-fi
-
-bun run lint:md
-md_status=$?
-if [ "$md_status" -ne 0 ]; then
-  echo "❌ pre-push: markdownlint reported violations — fix before pushing (or use --no-verify)" >&2
-  exit "$md_status"
-fi
-
-echo "✅ pre-push: lint gates passed (ESLint + markdownlint)"
-exit 0

package/.github/workflows/build-release-binaries.yml

@@ -1,163 +0,0 @@
-name: Build and Publish Release Binaries
-
-# Trigger when a release is published (not drafts).
-on:
-  release:
-    types:
-      - published
-  workflow_dispatch:
-    inputs:
-      release_tag:
-        description: 'Existing GitHub release tag to build and upload (for example v0.0.17)'
-        required: true
-        type: string
-      source_ref:
-        description: 'Optional source ref; defaults to the release tag'
-        required: false
-        default: ''
-        type: string
-
-permissions:
-  contents: write
-
-jobs:
-  build-binary:
-    name: Build ${{ matrix.target }}
-    runs-on: ${{ matrix.os }}
-    strategy:
-      # Build every platform even if one fails so we can see all failures.
-      fail-fast: false
-      matrix:
-        include:
-          - target: linux-x64
-            os: ubuntu-latest
-            platform: linux
-            arch: x64
-            bun-target: bun-linux-x64
-            binary: savant-code
-          - target: linux-arm64
-            # Cross-compile ARM64 from an x64 runner so we don't depend on the
-            # public-beta ARM GitHub runners.
-            os: ubuntu-latest
-            platform: linux
-            arch: arm64
-            bun-target: bun-linux-arm64
-            binary: savant-code
-          - target: darwin-x64
-            # Cross-compile Intel macOS from the available macOS runner.
-            os: macos-latest
-            platform: darwin
-            arch: x64
-            bun-target: bun-darwin-x64
-            binary: savant-code
-          - target: darwin-arm64
-            os: macos-latest
-            platform: darwin
-            arch: arm64
-            bun-target: bun-darwin-arm64
-            binary: savant-code
-          - target: win32-x64
-            os: windows-latest
-            platform: win32
-            arch: x64
-            bun-target: bun-windows-x64
-            binary: savant-code.exe
-
-    steps:
-      - name: Resolve release tag
-        id: release-tag
-        shell: bash
-        env:
-          INPUT_RELEASE_TAG: ${{ inputs.release_tag }}
-          INPUT_SOURCE_REF: ${{ inputs.source_ref }}
-          EVENT_RELEASE_TAG: ${{ github.event.release.tag_name }}
-        run: |
-          TAG="${INPUT_RELEASE_TAG:-$EVENT_RELEASE_TAG}"
-          if [ -z "$TAG" ]; then
-            echo "A release tag is required." >&2
-            exit 1
-          fi
-          echo "TAG=$TAG" >> "$GITHUB_OUTPUT"
-          SOURCE_REF="${INPUT_SOURCE_REF:-$TAG}"
-          echo "SOURCE_REF=$SOURCE_REF" >> "$GITHUB_OUTPUT"
-
-      - name: Checkout release source
-        uses: actions/checkout@v4
-        with:
-          ref: ${{ steps.release-tag.outputs.SOURCE_REF }}
-
-      - name: Setup Bun
-        uses: oven-sh/setup-bun@v1
-        with:
-          bun-version: 1.3.14
-
-      - name: Restore dependencies cache
-        id: cache-node-modules
-        uses: actions/cache@v4
-        with:
-          path: |
-            **/node_modules
-            ~/.bun/install/cache
-          key: bun-${{ runner.os }}-${{ hashFiles('**/bun.lock', '**/bun.lockb', 'package.json') }}
-          restore-keys: |
-            bun-${{ runner.os }}-
-
-      - name: Install dependencies
-        run: bun install --frozen-lockfile
-
-      - name: Extract version from tag
-        id: version
-        shell: bash
-        env:
-          RELEASE_TAG: ${{ steps.release-tag.outputs.TAG }}
-        run: |
-          echo "VERSION=${RELEASE_TAG#v}" >> "$GITHUB_OUTPUT"
-
-      - name: Build binary
-        working-directory: cli
-        env:
-          # Release binaries must run in production mode. SAVANT_CODE_BUILD_ENV
-          # lets dev builds opt out; in CI we always want prod.
-          NEXT_PUBLIC_CB_ENVIRONMENT: prod
-          NEXT_PUBLIC_SAVANT_CODE_APP_URL: https://savant-code.com
-          NEXT_PUBLIC_WEB_PORT: 3000
-          NEXT_PUBLIC_SUPPORT_EMAIL: support@savant-code.com
-          NEXT_PUBLIC_POSTHOG_API_KEY: phc_release_placeholder
-          NEXT_PUBLIC_POSTHOG_HOST_URL: https://us.i.posthog.com
-          NEXT_PUBLIC_GRAVITY_PIXEL_ID: 00000000-0000-0000-0000-000000000000
-          NEXT_PUBLIC_STRIPE_PUBLISHABLE_KEY: pk_release_placeholder
-          NEXT_PUBLIC_STRIPE_CUSTOMER_PORTAL: https://savant-code.com/portal
-          NEXT_PUBLIC_GOOGLE_SITE_VERIFICATION_ID: release_placeholder
-          # Pin the build target so the native runner architecture does not
-          # accidentally influence which binary is produced.
-          OVERRIDE_TARGET: ${{ matrix.bun-target }}
-          OVERRIDE_PLATFORM: ${{ matrix.platform }}
-          OVERRIDE_ARCH: ${{ matrix.arch }}
-        run: |
-          bun run scripts/build-binary.ts savant-code "${{ steps.version.outputs.VERSION }}"
-
-      - name: Smoke test binary
-        # ARM64 and cross-compiled Intel macOS binaries cannot be assumed runnable
-        # on the runner architecture; the build and packaged asset remain validated.
-        if: matrix.arch != 'arm64' && matrix.target != 'darwin-x64'
-        shell: bash
-        run: |
-          ./cli/bin/${{ matrix.binary }} --version
-
-      - name: Package tarball
-        shell: bash
-        run: |
-          cd cli/bin
-          tar -czf "savant-code-${{ matrix.target }}.tar.gz" "${{ matrix.binary }}" tree-sitter.wasm env.json
-
-      - name: Upload to GitHub release
-        shell: bash
-        env:
-          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        run: |
-          gh release upload "${{ steps.release-tag.outputs.TAG }}" "cli/bin/savant-code-${{ matrix.target }}.tar.gz" --clobber
-
-# Note: baseline AVX2 fallbacks (linux-x64-baseline, win32-x64-baseline) are
-# not yet built because the current build script does not produce them. Add a
-# baseline build step once Bun supports a baseline target flag or the build
-# script is extended to emit non-AVX2 binaries.

package/.markdownlint.json

@@ -1,42 +0,0 @@
-{
-  "default": false,
-  "MD001": true,
-  "MD003": { "style": "atx" },
-  "MD012": true,
-  "MD013": {
-    "line_length": 120,
-    "heading_line_length": 120,
-    "code_blocks": false,
-    "tables": false
-  },
-  "MD022": true,
-  "MD032": true,
-  "MD033": {
-    "allowed_elements": [
-      "a",
-      "b",
-      "br",
-      "code",
-      "details",
-      "div",
-      "em",
-      "i",
-      "img",
-      "kbd",
-      "p",
-      "span",
-      "strong",
-      "sub",
-      "summary",
-      "sup",
-      "table",
-      "tbody",
-      "td",
-      "th",
-      "thead",
-      "tr"
-    ]
-  },
-  "MD040": true,
-  "MD041": true
-}

package/.markdownlintignore

@@ -1,21 +0,0 @@
-node_modules/
-dev/fids/archive/**
-CHANGELOG.md
-research/
-resources/
-dev/test-prompts/archive/**
-dev/nova/inbox/archive/**
-dev/nova/outbox/archive/**
-dev/nova/reports/**
-docs/cloudflare-llms-full.md
-docs/Savant-Code Business And Backend Research.md
-docs/Savant Code Launch Strategy.md
-docs/reports/Thinker Agent Architecture Research.md
-docs/reports/Savant-Code Benchmark Specification.md
-docs/reports/repos/**
-docs/reports/feature-parity-report.md
-dev/archive/2026-08-02-repository-hygiene/**
-docs/reports/codebuff-discord-feedback.md
-docs/reports/**
-docs/launch/**
-docs/Launch Plan Review and Optimization.md

package/.prettierignore

@@ -1,9 +0,0 @@
-# Markdown formatting is owned by markdownlint (`bun run lint:md` — see .markdownlint.json).
-# Prettier's markdown defaults (80-char prose wrap, padded tables) conflict with the repo's
-# 120-char-line / compact-table style, so markdown is excluded from the format gate.
-*.md
-
-# Local content dirs (already gitignored; listed here so prettier skips them explicitly)
-resources/
-research/
-.eslint-recovery.json

package/.prettierrc

@@ -1,23 +0,0 @@
-{
-  "tabWidth": 2,
-  "useTabs": false,
-  "semi": false,
-  "trailingComma": "all",
-  "singleQuote": true,
-  "overrides": [
-    {
-      "files": "*.sql",
-      "options": {
-        "language": "postgresql",
-        "keywordCase": "lower",
-        "logicalOperatorNewline": "before"
-      }
-    },
-    {
-      "files": "*.svg",
-      "options": {
-        "parser": "html"
-      }
-    }
-  ]
-}

package/.savantignore

@@ -1,3 +0,0 @@
-!codebuff.json
-__mock-projects__
-test-repos

package/AGENTS.md

@@ -1,97 +0,0 @@
-# Savant-Code
-
-Savant-Code is the multi-agent AI coding assistant built on the ECHO Protocol v0.2.0.
-
-> **All work is governed by [ECHO Protocol](ECHO.md).** No code is written without a FID that has converged through the
-  Perfection Loop.
-
-## Key Technologies
-
-- **TypeScript monorepo** (`strict: true`, `noImplicitReturns: true`)
-- **Bun runtime + package manager** (≥ 1.3.11; root pins `1.3.14` per `engines.bun`)
-- **OpenTUI + React CLI** (`@opentui/core` 0.2.2, `react` ^19)
-- **JS/TS SDK** — embeddable via `SavantClient`
-- **Composable agent runtime** — ECHO 9-agent roster with separation of duties
-- **ECHO Protocol v0.2.0** — single bootstrap file
-
-## Agent Roster (9 canonical ECHO roles)
-
-The 9-agent roster is enforced in `ARCHITECTURE.md`. Each agent has restricted tools. See
-[ARCHITECTURE.md](ARCHITECTURE.md) → "Helper Tool Libraries" for the distinction between the 9 runtime roles and the 14
-filesystem entries in `agents/`.
-
-| Agent | Phase | Role |
-|---|---|---|
-| Orchestrator | ALL | Routes work, enforces protocol, spawns agents |
-| Detective | RED | Codebase analysis + grep call-graphs |
-| Forge | GREEN | Implementation only — writes code from converged FID |
-| Verifier | AUDIT | Double-audit, run tests, check call-graph |
-| Recorder | FID | Create, track, archive FIDs. Update CHANGELOG. |
-| Thinker | Planning | Sequential reasoning via `sequentialthinking` tool |
-| Scout | Explore | Glob/list_directory/read_files for context gathering |
-| Researcher | Research | Web search + docs lookup |
-| Scribe | Docs | Session summaries, LESSONS.md, knowledge files |
-
-## Repo Map
-
-| Workspace | Package | Purpose |
-|---|---|---|
-| `agents/` | `@savant-code/agents` | Public agent definitions shipped with the CLI (9 canonical + 6 helper tool-library/trace dirs) |
-| `cli/` | `@savant-code/cli` | CLI source — UI, commands, state, hooks, OpenTUI/React components |
-| `common/` | `@savant-code/common` | Shared types, tool definitions, utilities |
-| `evals/` | `@savant-code/evals` | Eval benchmark runner + public eval fixtures |
-| `savant-free/` | `@savant-code/savant-free` | CLI release + e2e tests for the free variant |
-| `packages/database/` | `@savant-code/database` | Database abstraction layer |
-| `packages/agent-runtime/` | `@savant-code/agent-runtime` | Agent loop, tool executor, LLM API integration |
-| `packages/code-map/` | `@savant-code/code-map` | tree-sitter code indexing, language detection |
-| `packages/llm-providers/` | `@savant-code/llm-providers` | Public LLM provider shims |
-| `sdk/` | `@savant-code/sdk` | Public SDK — `SavantClient`, types, build + verify scripts |
-
-## Conventions
-
-- Use `bun install` and `bun run`.
-- Prefer dependency injection over module mocking.
-- Run interactive CLI tests in tmux.
-- Do not force-push `main`.
-- All typecheck × 4 must pass before any merge: `sdk`, `common`, `packages/agent-runtime`, `cli`.
-- New FIDs follow the format `dev/fids/FID-YYYY-MMDD-NNN-{title}.md`.
-
-## Docs
-
-- [**ECHO.md**](ECHO.md) — The 15 Laws + Perfection Loop FSM + FID lifecycle *(READ 0-EOF FIRST)*
-- [**ARCHITECTURE.md**](ARCHITECTURE.md) — Agent roster and tool restrictions
-- [**CONTRIBUTING.md**](CONTRIBUTING.md) — This file (ECHO Protocol contributor guide)
-- [**protocol.config.yaml**](protocol.config.yaml) — Build commands, quality bar, paths
-- [**CHANGELOG.md**](CHANGELOG.md) — Reverse-chronological release history (FID-archived entries)
-- [**dev/fids/**](dev/fids/) — Active FIDs awaiting resolution
-- [**dev/fids/archive/**](dev/fids/archive/) — Closed FIDs (per Auto-Archive rule)
-- [**dev/LEARNINGS.md**](dev/LEARNINGS.md) — Cross-session lessons
-- [**dev/session-summaries/**](dev/session-summaries/) — Session audit trail
-- [**dev/test-prompts/**](dev/test-prompts/) — A-Z test prompts and reports
-- [**dev/nova/inbox/**](dev/nova/inbox/) + [**outbox/**](dev/nova/outbox/) — Third-party audit channel
-- [**dev/scratchpad/**](dev/scratchpad/) — Ephemeral working area (Orchestrator writes via `/dev override`)
-- [**`.agents/skills/`**](.agents/skills/) — 7 coding standards as `SKILL.md` skills (auto-loaded)
-
-## Validation
-
-```bash
-# Typecheck × 4 (HARD GATE)
-cd sdk && bun run typecheck && cd ../common && bun run typecheck && cd ../packages/agent-runtime && bun run typecheck && cd ../../cli && bun run typecheck
-
-# SDK test suite (HARD GATE)
-cd sdk && bun test src/
-
-# Lint (HARD GATE — enforced pre-push)
-bun x eslint . --max-warnings 0
-
-# Markdown lint (HARD GATE — enforced pre-push)
-bun run lint:md
-
-# Format
-bun x prettier --write .
-```
-
-> **Pre-push gate:** a native git pre-push hook (`.githooks/pre-push`) runs
-> `bun x eslint . --max-warnings 0` + `bun run lint:md` before every push. It is
-> auto-wired on `bun install` via the root `prepare` script
-> (`git config core.hooksPath .githooks`). Bypass only with `git push --no-verify`.