Corpus inventory

SKILL: Corpus Inventory Wiki

Root Directory: corpus_inventory_wiki_rag_vector/ All paths below are relative to that root.

Purpose

This wiki is a model-agnostic RAG index — any agent from any provider can read these markdown files to understand what books and documents exist in the 0sXai corpus, where they're located, and what topics they cover. No embeddings, no vector DB — just structured markdown that any LLM can parse.

Execution Flow

1. Boot: Read system_directive.md → knowledge/skill_context.md → this file.
2. Orient: Read wiki/index.md for the full corpus map, then wiki/hot.md for recent changes.
3. Query: Navigate to the relevant category page in wiki/categories/ for detailed book listings.
4. Report: If you discover new files or reclassifications, append to working_doc.md.

Wiki Structure

corpus_inventory_wiki_rag_vector/
├── system_directive.md              # Governance (read FIRST)
├── SANDBOX_RULES.md                 # Project rules
├── working_doc.md                   # Living changelog + recommendations
├── knowledge/
│   └── skill_context.md             # Environment map (read SECOND)
├── skills/
│   └── corpus-inventory/
│       └── SKILL.md                 # This file (read THIRD)
├── wiki/
│   ├── index.md                     # Master catalog — entry point for all agents
│   ├── hot.md                       # Recent changes cache (~500 tokens)
│   ├── corpus_health.md             # Dedup report, gap analysis, metrics
│   └── categories/                  # One page per topic domain
│       ├── architecture.md
│       ├── agile_planning.md
│       ├── software_dev.md
│       ├── ai_ml.md
│       ├── blockchain_defi.md
│       ├── security.md
│       ├── math_physics.md
│       ├── cloud_infra.md
│       ├── data_engineering.md
│       ├── quant_trading.md
│       └── uncategorized.md
├── inventory/                       # Machine-readable data
│   ├── full_inventory.json          # Complete deduped inventory
│   └── classification_report.json   # Heuristic results per file
├── session_logs/                    # Operational diary
│   └── session_log_YYYY-MM-DD.md
└── scripts/                         # Inventory generation scripts
    └── build_inventory.py

Category Taxonomy

ID	Category	Keywords	Agent Use Cases
CAT-01	Architecture	architect, design pattern, microservice, system design, clean code, DDD, distributed	System design agents, code review agents
CAT-02	Agile/Planning	agile, scrum, kanban, sprint, lean, project management, devops	Project planning agents, sprint orchestrators
CAT-03	Software Dev	programming, python, javascript, react, flutter, web dev, API, full-stack	Coding agents, development assistants
CAT-04	AI/ML	artificial intelligence, machine learning, deep learning, neural, NLP, LLM, transformer	Research agents, ML pipeline agents
CAT-05	Blockchain/DeFi	blockchain, solidity, smart contract, defi, crypto, bitcoin, web3, dao	DeFi agents, smart contract auditors
CAT-06	Security	security, cybersecurity, hacking, penetration, zero trust, red team, OWASP	Security auditor agents, red team agents
CAT-07	Math/Physics	mathematics, calculus, algebra, statistics, probability, physics, quantum, topology	Research agents, quant agents
CAT-08	Cloud/Infra	aws, azure, gcp, kubernetes, docker, terraform, infrastructure, serverless	DevOps agents, infrastructure agents
CAT-09	Data Engineering	data science, database, sql, spark, kafka, airflow, etl, analytics	Data pipeline agents, analytics agents
CAT-10	Quant/Trading	<sanitized_trading_doc_1>	Trading bot agents, quant researchers
CAT-99	Uncategorized	(no match)	Flagged for human review

Must Do

• ALWAYS read wiki/index.md before answering corpus questions.
• ALWAYS cite the category page and book entry when recommending a source.
• ALWAYS append discoveries to working_doc.md.
• ALWAYS use Obsidian [[wikilinks]] for cross-references.

Must NOT Do

• NEVER modify source files on disk.
• NEVER guess book contents from titles alone — use the heuristic confidence score.
• NEVER merge categories without updating the taxonomy table above.
• NEVER skip the confidence level field — LOW is acceptable, missing is not.

Book Entry Format (Wiki Category Pages)

Each book in a category page follows this structure:

### [Book Title]
- **Author:** [Author name or "Unknown"]
- **Year:** [Publication year or "Unknown"]
- **Size:** [X.X MB]
- **Format:** [PDF/EPUB/MOBI]
- **Confidence:** [HIGH/MEDIUM/LOW]
- **Keywords:** [keyword1, keyword2, keyword3]
- **Heuristic Layers:** [L1-Filename, L3-Year, L5-Size]
- **Source Path:** `[canonical path]`
- **Copies Found:** [N] locations
- **Agent Relevance:** [Which agents should read this]
- **Cross-refs:** [[related_category]], [[related_book]]

Working Doc Update Protocol

After every session that touches this project, append to working_doc.md:

## Update: YYYY-MM-DD HH:MM TZ

### Changes
- [What was added/modified/reclassified]

### Corpus Health Snapshot
- Total unique files: [N]
- Classified: [N] | Unclassified: [N]
- Duplicates removed: [N]
- Categories with gaps: [list]

### Recommendations
- [Suggested fixes, reorganization, missing coverage]

### Open Questions
- [Questions that need human answers to proceed]