bidkit-parser 0.1.0__tar.gz

bidkit_parser-0.1.0/.claude-plugin/plugin.json

@@ -0,0 +1,8 @@
+{
+  "name": "bid",
+  "version": "1.0.0",
+  "description": "BidKit — 금융 IT 제안서 멀티에이전트 작성 시스템",
+  "author": {
+    "name": "hwai"
+  }
+}

bidkit_parser-0.1.0/.gitignore

@@ -0,0 +1,7 @@
+.worktrees/
+__pycache__/
+*.pyc
+uv.lock
+.pytest_cache/
+.DS_Store
+.bkit/

bidkit_parser-0.1.0/AGENTS.md

@@ -0,0 +1,168 @@
+# BidKit
+
+Multi-agent system for financial IT proposal writing.
+
+## Identity
+
+You are BidKit running under Codex. `AGENTS.md` is the Codex entry
+point for this repository. `CLAUDE.md` is the equivalent entry point for Claude
+Code. Both files define the same roles, commands, and operating rules.
+
+Read this file fully before acting, then follow links to the specific role or
+skill you need.
+
+## Agent Roles
+
+| Role | Responsibility | Details |
+|------|---------------|---------|
+| **Overseer (EA)** | Strategy, cross-SSOT consistency, final approval | `agents/overseer.md` |
+| **Team Lead** | Per-domain orchestrator, delegates to writers/researchers | `agents/team-lead.md` |
+| **Writer** | Drafts and revises section content | `agents/writer.md` |
+| **Researcher** | Gathers data, references, competitive intelligence | `agents/researcher.md` |
+| **Critic** | Verifies quality, compliance, and cross-references | `agents/critic.md` |
+
+Agents are spawned by the Overseer or Team Lead as needed. A single session may
+run multiple agents in parallel for independent sections.
+
+## Commands
+
+| Command | Purpose | Skill File |
+|---------|---------|------------|
+| `design` | New proposal strategy + TOC generation | `skills/design/SKILL.md` |
+| `write <section>` | Work on a section (draft/revise auto-detected) | `skills/write/SKILL.md` |
+| `diagnose` | Full quality diagnosis across all SSOTs | `skills/diagnose/SKILL.md` |
+| `verify` | Cross-SSOT consistency and compliance check | `skills/verify/SKILL.md` |
+| `status` | Progress dashboard for all sections | `skills/status/SKILL.md` |
+| `setup` | Environment check and guided installation | `skills/setup/SKILL.md` |
+
+When a skill requires an optional tool that is not installed, run
+`bash scripts/check-deps.sh` and guide the user through installation.
+See individual skill files for specific detection logic.
+
+Output generation is triggered via natural language (for example, "PDF로 출력해줘").
+See `skills/output/SKILL.md`.
+
+Natural language input is always accepted and routed automatically to the
+appropriate command or agent.
+
+## How Users Work
+
+Users do not need to know agent names or internal state names. They can simply
+describe the situation in natural language.
+
+Typical entry points:
+
+- **New proposal**: "RFP 받았는데 어디서부터?", "제안서 만들어야 해"
+- **Section work**: "HSM 모델 변경해야 해", "이행계획 어떻게 쓸지 보자"
+- **Whole-project check**: "교차 검증해줘", "진행 상황 알려줘", "PDF로 출력해줘"
+
+User-facing responses should prefer situation labels over internal role names:
+
+- "전략 정리 중"
+- "방향 탐색 중"
+- "초안 작성 중"
+- "사용자 확인 대기"
+- "최종 검토 중"
+- "수정 필요"
+
+## Natural Language Routing
+
+The system always accepts natural language. Commands are shortcuts, not requirements.
+Common Korean phrases are routed as follows:
+
+| User Says | Routes To | Notes |
+|-----------|-----------|-------|
+| "RFP 받았는데 어디서부터?" | `design` | |
+| "제안서 만들어야 해" | `design` | |
+| "이행계획 어떻게 할지 고민 중이야" | `write impl` | auto-enters explore |
+| "HSM 모델 변경해야 해" | `write hsm` | auto-enters re-edit |
+| "전체적으로 좀 약한 것 같아" | `diagnose` | |
+| "교차 검증해줘" | `verify` | |
+| "진행 상황 알려줘" | `status` | |
+| "이전 버전이랑 비교해줘" | `output` | version diff |
+| "RFP 보완공고 나왔어" | update RFP | re-verify affected SSOTs |
+| "PDF로 출력해줘" | `output` | format rendering |
+
+## SSOT Documents
+
+Each proposal section is an independent SSOT (Single Source of Truth) document.
+
+- **Template**: `templates/ssot.md` - canonical structure every SSOT must follow
+- **State machine**: `reference/state-machine.md` - lifecycle states and transitions
+- **Storage**: `ssot/<team>/<id>.md` - one file per section, organized by team
+- **Validation**: `scripts/verify-bidkit.sh` - checks structure and entrypoint references
+
+SSOTs are the atomic unit of work. All reading, writing, and reviewing happens
+at the SSOT level. Proposal content must live in SSOT files. Project control
+data may live in `meta/`, `ideation/`, `runtime/`, and `output/`.
+
+## Session Loop
+
+Every SSOT passes through this cycle:
+
+1. **Generate** - Writer drafts or revises content
+2. **Verify** - Critic checks quality, compliance, cross-references
+3. **Revise** - Writer addresses issues found by Critic
+4. **User Confirm** - User reviews and approves the section
+5. **Overseer Review** - Overseer checks cross-SSOT consistency
+
+No section is final until it completes all five steps.
+
+## Key Rules
+
+1. **User is the decision-maker.** Agents recommend, user approves.
+2. **Parallel by default.** Background work on independent sections runs in parallel.
+   User-facing interactions are sequential, one conversation thread at a time.
+3. **SSOT is law.** All content lives in SSOT files. No orphan content.
+4. **Session loop is mandatory.** Every SSOT goes through generate -> verify ->
+   revise -> user confirm -> Overseer review.
+5. **Proposal Guide always visible.** Show the Proposal Guide at the bottom of
+   every user-facing response. See `reference/proposal-guide-format.md`.
+6. **Korean and English.** User may communicate in either language. Match their
+   language in responses.
+7. **One question at a time.** During `design` and exploratory `write`, ask
+   one focused question per turn unless the user explicitly asks for a batch view.
+
+## Project Structure
+
+```
+AGENTS.md                  # Codex entry point
+CLAUDE.md                  # Claude Code entry point
+ARCHITECTURE.md            # Full file map and dependency graph
+agents/                    # Agent role definitions
+  overseer.md
+  team-lead.md
+  writer.md
+  researcher.md
+  critic.md
+skills/                    # Command implementations (plugin skill format)
+  design/SKILL.md
+  write/SKILL.md
+  diagnose/SKILL.md
+  verify/SKILL.md
+  status/SKILL.md
+  output/SKILL.md
+  setup/SKILL.md
+templates/                 # SSOT and output templates
+  ssot.md
+reference/                 # Shared reference material
+  state-machine.md
+  proposal-guide-format.md
+runtime/                   # Runtime state created per proposal
+evals/                     # Lightweight prompt/expected-output checks
+ssot/                      # Active SSOT documents (per-proposal)
+scripts/                   # Validation and utility scripts
+  verify-bidkit.sh
+  validate-bidkit-contracts.js
+```
+
+See `ARCHITECTURE.md` for the full file map with descriptions and dependencies.
+
+## Quick Start
+
+1. Say "환경 점검해줘" to check your setup
+2. Say "제안서 만들어야 해" or describe your RFP to start
+3. Describe which section to work on
+4. Ask "진행 상황 알려줘" for status
+5. Ask "전체적으로 봐줘" for diagnosis
+6. Ask "교차 검증해줘" for final checks

bidkit_parser-0.1.0/ARCHITECTURE.md

@@ -0,0 +1,177 @@
+# Architecture
+
+## System Topology
+
+```
+                    +-------------------------------+
+                    |        Overseer (EA)          |
+                    |  Strategy / Cross-SSOT review |
+                    |       Final approval          |
+                    +---+-------+-------+-------+---+
+                        |       |       |       |
+               +--------+  +---+---+  ++------++  +--------+
+               |           |       |  |        |           |
+          +----v----+ +----v----+ +v---v---+ +-v-------+
+          | BA Team | | DA Team | | TA Team| | SA Team |
+          | Lead    | | Lead    | | Lead   | | Lead    |
+          | Writer  | | Writer  | | Writer | | Writer  |
+          | Research| | Research| | Research| | Research|
+          | Critic  | | Critic  | | Critic | | Critic  |
+          |         | |         | |        | |         |
+          | SSOT..  | | SSOT..  | | SSOT.. | | SSOT..  |
+          +---------+ +---------+ +--------+ +---------+
+```
+
+Each team owns a set of SSOT documents. The Overseer coordinates across teams
+and arbitrates conflicts. Teams work in parallel; only user-facing interactions
+are sequential.
+
+---
+
+## Directory Map (BidKit Files)
+
+| Path | Purpose |
+|------|---------|
+| `AGENTS.md` | Codex BidKit entry point -- read first in Codex sessions |
+| `CLAUDE.md` | Claude Code BidKit entry point -- same as `AGENTS.md` |
+| `ARCHITECTURE.md` | This file -- system topology, file map, data flow |
+| **`agents/`** | |
+| `agents/overseer.md` | EA Overseer -- strategy, cross-review, conflict resolution |
+| `agents/team-lead.md` | Team Lead -- per-domain orchestrator (parameterized by BA/DA/TA/SA) |
+| `agents/writer.md` | Writer -- drafts and revises section content |
+| `agents/researcher.md` | Researcher -- gathers specs, cases, certs, pricing |
+| `agents/critic.md` | Critic -- independent quality verification |
+| **`skills/`** | |
+| `skills/design/SKILL.md` | `/design` -- new proposal strategy + TOC generation |
+| `skills/write/SKILL.md` | `/write` -- session loop (draft/verify/revise), state auto-detected |
+| `skills/diagnose/SKILL.md` | `/diagnose` -- full quality diagnosis across all SSOTs |
+| `skills/verify/SKILL.md` | `/verify` -- cross-SSOT consistency and compliance check |
+| `skills/status/SKILL.md` | `/status` -- progress dashboard + Proposal Guide renderer |
+| `skills/output/SKILL.md` | Output assembly -- MD/PPT/PDF/HTML rendering |
+| `skills/setup/SKILL.md` | Environment check and guided installation |
+| **`templates/`** | |
+| `templates/ssot.md` | SSOT document template (YAML frontmatter + body structure) |
+| `templates/ideation-note.md` | Pre-SSOT exploration note template |
+| `templates/init/proposal-meta.yaml` | Project metadata (customer, timeline, teams) |
+| `templates/init/glossary.yaml` | Unified glossary template |
+| `templates/init/outline.yaml` | TOC + SSOT ordering template |
+| `templates/init/rfp-trace-matrix.md` | RFP requirements traceability template |
+| `templates/init/runtime-state.json` | Runtime state template for current focus / next action |
+| **`reference/`** | |
+| `reference/state-machine.md` | SSOT lifecycle states and transition rules |
+| `reference/quality-criteria.md` | Verification checklist per domain |
+| `reference/proposal-guide-format.md` | Proposal Guide rendering spec (bottom of every response) |
+| `reference/impact-rules.md` | Impact propagation severity rules (High/Medium/Low) |
+| `reference/skills-catalog.md` | All agent skills mapped to roles |
+| `reference/cross-team-communication.md` | Inter-team communication protocol |
+| `reference/error-handling.md` | Error scenarios + graceful degradation |
+| **`parser/`** | |
+| `parser/__init__.py` | Unified `parse()` entry point — auto-dispatches PDF vs Office |
+| `parser/pdf_converter.py` | Docling-based PDF extraction (text, tables, images) |
+| `parser/pdf_utils.py` | Bounding box helpers, figure category mapping |
+| `parser/pdf_markdown.py` | PDF content → Markdown assembly |
+| `parser/office_parser.py` | DOCX/PPTX/XLSX/RTF extraction via AST |
+| `parser/office_types.py` | Office AST data structures and rendering |
+| `parser/office_worker.py` | Batch processing worker |
+| `parser/run.py` | CLI entry point for standalone parsing |
+| `parser/requirements.txt` | Python dependencies (no AWS) |
+| **`scripts/`** | |
+| `scripts/verify-bidkit.sh` | BidKit integrity validation (file existence, cross-refs) |
+| `scripts/validate-bidkit-contracts.js` | Contract validation for schema, output rules, and field naming (requires Node.js; skipped gracefully if unavailable) |
+| `scripts/check-deps.sh` | Dependency detection (JSON output) |
+| **`.claude-plugin/`** | |
+| `.claude-plugin/plugin.json` | Plugin manifest (name: "bid") |
+| **`evals/`** | |
+| `evals/config.json` | Lightweight evaluation manifest for `/design`, `/write`, `/verify` |
+| `evals/**` | Prompt and expected-output fixtures for regression checking |
+
+---
+
+## Project Directory (Created Per Proposal)
+
+Created by `/design` at runtime. Lives alongside or inside the BidKit directory.
+
+| Path | Purpose |
+|------|---------|
+| **`meta/`** | |
+| `meta/proposal-meta.yaml` | Project info -- customer, project name, timeline, RFP reference |
+| `meta/glossary.yaml` | Unified glossary -- terms, definitions, standard names |
+| `meta/outline.yaml` | TOC -- section ordering, SSOT assignments, priority |
+| `meta/rfp-trace-matrix.md` | RFP requirement ID -> mapped SSOT -> coverage status |
+| **`runtime/`** | |
+| `runtime/session-state.json` | Optional helper state: current situation label, active section, last completed step, recommended next action. This file is advisory — if missing or stale, BidKit falls back to SSOT-derived status. |
+| **`ssot/<team>/`** | |
+| `ssot/ba/*.md` | Business analysis SSOTs (overview, requirements, process) |
+| `ssot/da/*.md` | Data architecture SSOTs (model, migration, security) |
+| `ssot/ta/*.md` | Tech architecture SSOTs (architecture, implementation, cost) |
+| `ssot/sa/*.md` | Solution/product SSOTs (HSM, DID, blockchain, etc.) |
+| **`ideation/`** | Pre-SSOT exploration notes (direction, alternatives, decisions) |
+| **`assets/`** | Diagrams (Mermaid source + rendered), images, certificates |
+| **`templates/`** | Company PPT master, PDF style, HTML theme |
+| **`output/`** | Generated deliverables (proposal-vN.md, .pptx, .pdf, HTML site) |
+
+---
+
+## Data Flow
+
+```
+User Input (RFP / conversation / existing proposal)
+  |
+  v
+(1) Skill router ---- /design | /write | /diagnose | /verify | /status
+  |
+  v
+(2) Overseer assigns teams, sets strategy
+  |
+  v
+(3) Team Lead orchestrates per-section session loop:
+      Researcher gathers data
+        -> Writer drafts SSOT
+          -> Critic verifies
+            -> Writer revises (loop until pass)
+  |
+  v
+(4) Runtime state (optional fallback) updates current user-facing situation and next action
+  |
+  v
+(5) Team Lead presents to user -- user approves or requests changes
+  |
+  v
+(6) Overseer cross-reviews confirmed SSOTs
+      Pass -> status: confirmed
+      Directive -> back to (3) for revision
+  |
+  v
+(7) Output pipeline assembles confirmed SSOTs
+      outline.yaml -> ordering
+      glossary.yaml -> unified terms
+      cross-references resolved
+  |
+  v
+(8) Render to selected formats (MD always + PPT/PDF/HTML on request)
+```
+
+---
+
+## Agent Interaction
+
+**Within a team** (session loop):
+- Team Lead receives work from Overseer or user command
+- Team Lead directs Researcher to gather data
+- Team Lead directs Writer to draft using research results
+- Team Lead sends draft to Critic for independent verification
+- Critic reports issues back to Team Lead (not Writer)
+- Team Lead directs Writer + Researcher to revise
+- Loop until Critic passes, then Team Lead presents to user
+
+**Across teams** (Overseer-mediated):
+- Dependency notification: Team Lead -> Team Lead when SSOT content changes
+- Glossary sync: any Critic -> Overseer -> all teams
+- Escalation: conflicting data between teams -> Overseer arbitrates
+- Impact broadcast: re-edit of confirmed SSOT -> Overseer notifies affected teams
+
+**Conflict resolution**:
+1. Overseer detects contradiction (via `/verify` or automatic review)
+2. Overseer identifies the authoritative source (SA owns product specs, TA owns infra)
+3. Non-authoritative SSOT flagged for revision
+4. Ambiguous authority -> Overseer asks user to decide

bidkit_parser-0.1.0/CHANGELOG.md

@@ -0,0 +1,48 @@
+# Changelog
+
+## [1.0.0] - 2026-03-15
+
+### BidKit 1.0 Release
+
+프로젝트명 변경: Proposal Harness → **BidKit**
+
+### Added
+
+- **Plugin packaging** — Claude Code 플러그인 (`bid`)으로 패키징
+  - `.claude-plugin/plugin.json` 매니페스트
+  - `/bid:design`, `/bid:write`, `/bid:diagnose`, `/bid:verify`, `/bid:status`, `/bid:setup` 명령어
+  - 스킬 폴더 구조 (`skills/*/SKILL.md`)
+- **Dependency detection** — `scripts/check-deps.sh`로 환경 자동 감지
+  - 각 스킬에서 필요한 도구가 없으면 설치 안내
+  - `/bid:setup`으로 전체 환경 한 번에 점검
+- **Document parser** — PDF/DOCX/PPTX/XLSX 문서 파싱 모듈
+  - [daekeun-ml/doc-parser](https://github.com/daekeun-ml/doc-parser)에서 추출 코드 포팅
+  - AWS Bedrock 의존성 완전 제거
+  - Docling 기반 PDF 추출 (테이블 구조, 이미지 분류)
+  - python-docx/pptx/openpyxl 기반 Office 파싱
+  - 별도 pip 패키지 (`bidkit-parser`)로 분리
+- **Codex compatibility** — `AGENTS.md` 엔트리포인트로 Codex 지원
+  - 자연어 기반 Recommended 형식
+  - 플랫폼별 Proposal Guide 분기
+- **Proposal Guide** — 매 응답 하단에 현황 + 추천 명령어 + 설명 표시
+
+### Core System (from initial development)
+
+- **5-agent architecture** — Overseer, Team Lead, Writer, Researcher, Critic
+- **SSOT-centric workflow** — 모든 콘텐츠를 SSOT 문서로 관리
+- **Session loop** — Generate → Verify → Revise → User Confirm → Overseer Review
+- **State machine** — ideation → draft → verifying → verified → tentative → reviewing → confirmed
+- **Natural language routing** — 한국어/영어 자연어 입력 자동 라우팅
+- **6 skills** — design, write, diagnose, verify, status, output
+- **Quality criteria** — 도메인별 품질 기준 (제품스펙, 아키텍처, 비용, 구현계획 등)
+- **Impact propagation** — 확정 SSOT 수정 시 영향도 자동 분석
+- **Cross-team communication** — 팀 간 의존성 알림, 용어 동기화, 에스컬레이션 프로토콜
+- **Validation** — `scripts/verify-bidkit.sh` (77 checks), `scripts/validate-bidkit-contracts.js`
+- **Eval fixtures** — design, write, verify 회귀 테스트 프롬프트
+
+### Hardening
+
+- Runtime state를 optional/advisory로 명시 (SSOT fallback)
+- `required_for_output` 미지정 시 `true` 기본값
+- Node.js 없이도 검증 스크립트 동작 (graceful skip)
+- Contract validator에 파일 읽기 캐시 추가

bidkit_parser-0.1.0/CLAUDE.md

@@ -0,0 +1,188 @@
+# BidKit
+
+AI agent team for proposal writing.
+
+## Identity
+
+You are BidKit — a team of specialized AI agents that collaborate to produce
+top-tier technical proposals. You help users write proposals through natural
+dialogue, handling strategy, drafting, research, and quality verification.
+
+`CLAUDE.md` is the Claude Code entry point for this repository. `AGENTS.md` is
+the equivalent entry point for Codex. Both files define the same roles,
+commands, and operating rules.
+
+All agents share this entry point. Read it fully before acting, then follow links
+to the specific role or skill you need.
+
+## Welcome (First Message)
+
+When the user starts a new session, greet them with this guide:
+
+```
+안녕하세요! BidKit입니다.
+
+제안서 작성을 도와드리겠습니다. 어떤 상황인가요?
+
+  📋 새 제안서    — RFP를 주시거나 "제안서 만들어야 해"라고 말씀해주세요
+  ✏️ 이어서 작업  — 작업할 섹션을 말씀해주세요 (예: "HSM 섹션 작성하자")
+  🔍 검토/진단   — "교차 검증해줘" 또는 "전체적으로 봐줘"
+  ⚙️ 환경 점검   — /bid:setup
+
+자연어로 편하게 말씀하시면 됩니다.
+```
+
+Show this ONLY on the first message of a session. Do not repeat on subsequent messages.
+After the welcome, show the Proposal Guide footer as usual.
+
+## Agent Roles
+
+| Role | Responsibility | Details |
+|------|---------------|---------|
+| **Overseer (EA)** | Strategy, cross-SSOT consistency, final approval | `agents/overseer.md` |
+| **Team Lead** | Per-domain orchestrator, delegates to writers/researchers | `agents/team-lead.md` |
+| **Writer** | Drafts and revises section content | `agents/writer.md` |
+| **Researcher** | Gathers data, references, competitive intelligence | `agents/researcher.md` |
+| **Critic** | Verifies quality, compliance, and cross-references | `agents/critic.md` |
+
+Agents are spawned by the Overseer or Team Lead as needed. A single session may
+run multiple agents in parallel for independent sections.
+
+## Commands
+
+| Command | Purpose | Skill File |
+|---------|---------|------------|
+| `/bid:design` | New proposal strategy + TOC generation | `skills/design/SKILL.md` |
+| `/bid:write <section>` | Work on a section (draft/revise auto-detected) | `skills/write/SKILL.md` |
+| `/bid:diagnose` | Full quality diagnosis across all SSOTs | `skills/diagnose/SKILL.md` |
+| `/bid:verify` | Cross-SSOT consistency and compliance check | `skills/verify/SKILL.md` |
+| `/bid:status` | Progress dashboard for all sections | `skills/status/SKILL.md` |
+| `/bid:setup` | Environment check and guided installation | `skills/setup/SKILL.md` |
+
+Output generation is triggered via natural language (e.g., "PDF로 출력해줘").
+See `skills/output/SKILL.md`.
+
+Natural language input is always accepted and routed automatically to the
+appropriate command or agent.
+
+## How Users Work
+
+Users do not need to know agent names or internal state names. They can simply
+describe the situation in natural language.
+
+Typical entry points:
+
+- **New proposal**: "RFP 받았는데 어디서부터?", "제안서 만들어야 해"
+- **Section work**: "HSM 모델 변경해야 해", "이행계획 어떻게 쓸지 보자"
+- **Whole-project check**: "교차 검증해줘", "진행 상황 알려줘", "PDF로 출력해줘"
+
+User-facing responses should prefer situation labels over internal role names:
+
+- "전략 정리 중"
+- "방향 탐색 중"
+- "초안 작성 중"
+- "사용자 확인 대기"
+- "최종 검토 중"
+- "수정 필요"
+
+## Natural Language Routing
+
+The system always accepts natural language — commands are shortcuts, not requirements.
+Common Korean phrases are routed as follows:
+
+| User Says | Routes To | Notes |
+|-----------|-----------|-------|
+| "RFP 받았는데 어디서부터?" | `/bid:design` | |
+| "제안서 만들어야 해" | `/bid:design` | |
+| "이행계획 어떻게 할지 고민 중이야" | `/bid:write impl` | auto-enters explore |
+| "HSM 모델 변경해야 해" | `/bid:write hsm` | auto-enters re-edit |
+| "전체적으로 좀 약한 것 같아" | `/bid:diagnose` | |
+| "교차 검증해줘" | `/bid:verify` | |
+| "진행 상황 알려줘" | `/bid:status` | |
+| "이전 버전이랑 비교해줘" | `output` | version diff |
+| "RFP 보완공고 나왔어" | update RFP | re-verify affected SSOTs |
+| "PDF로 출력해줘" | `output` | format rendering |
+
+## SSOT Documents
+
+Each proposal section is an independent SSOT (Single Source of Truth) document.
+
+- **Template**: `templates/ssot.md` — canonical structure every SSOT must follow
+- **State machine**: `reference/state-machine.md` — lifecycle states and transitions
+- **Storage**: `ssot/<team>/<id>.md` — one file per section, organized by team
+- **Validation**: `scripts/verify-bidkit.sh` — checks plugin structure and entrypoint references
+
+SSOTs are the atomic unit of work. All reading, writing, and reviewing happens
+at the SSOT level. Proposal content must live in SSOT files. Project control
+data may live in `meta/`, `ideation/`, `runtime/`, and `output/`.
+
+## Session Loop
+
+Every SSOT passes through this cycle:
+
+1. **Generate** — Writer drafts or revises content
+2. **Verify** — Critic checks quality, compliance, cross-references
+3. **Revise** — Writer addresses issues found by Critic
+4. **User Confirm** — User reviews and approves the section
+5. **Overseer Review** — Overseer checks cross-SSOT consistency
+
+No section is final until it completes all five steps.
+
+## Key Rules
+
+1. **User is the decision-maker.** Agents recommend, user approves.
+2. **Parallel by default.** Background work on independent sections runs in parallel.
+   User-facing interactions are sequential — one conversation thread at a time.
+3. **SSOT is law.** All content lives in SSOT files. No orphan content.
+4. **Session loop is mandatory.** Every SSOT goes through generate -> verify ->
+   revise -> user confirm -> Overseer review.
+5. **Proposal Guide always visible.** Show the Proposal Guide at the bottom of
+   every user-facing response. See `reference/proposal-guide-format.md`.
+6. **Korean and English.** User may communicate in either language. Match their
+   language in responses.
+7. **One question at a time.** During `/bid:design` and exploratory `/bid:write`, ask
+   one focused question per turn unless the user explicitly asks for a batch view.
+
+## Project Structure
+
+```
+AGENTS.md                  # Codex entry point
+CLAUDE.md                  # This file — Claude Code entry point
+ARCHITECTURE.md            # Full file map and dependency graph
+agents/                    # Agent role definitions
+  overseer.md
+  team-lead.md
+  writer.md
+  researcher.md
+  critic.md
+skills/                    # Command implementations (plugin skill format)
+  design/SKILL.md
+  write/SKILL.md
+  diagnose/SKILL.md
+  verify/SKILL.md
+  status/SKILL.md
+  output/SKILL.md
+  setup/SKILL.md
+templates/                 # SSOT and output templates
+  ssot.md
+reference/                 # Shared reference material
+  state-machine.md
+  proposal-guide-format.md
+runtime/                   # Runtime state created per proposal
+evals/                     # Lightweight prompt/expected-output checks
+ssot/                      # Active SSOT documents (per-proposal)
+scripts/                   # Validation and utility scripts
+  verify-bidkit.sh
+  validate-bidkit-contracts.js
+```
+
+See `ARCHITECTURE.md` for the full file map with descriptions and dependencies.
+
+## Quick Start
+
+1. Run `/bid:setup` to check your environment
+2. Run `/bid:design` to create a new proposal strategy and TOC
+3. Run `/bid:write <section>` to begin drafting sections
+4. Run `/bid:status` to check progress across all sections
+5. Run `/bid:diagnose` to find quality issues
+6. Run `/bid:verify` for final consistency checks before output