knowledge-compiler
Installation
SKILL.md
Knowledge Compiler
Agents do research, pull data, write reports—the wiki connects these outputs. The more you use the agent, the more it understands your domain.
Runtime Dependencies & Permissions
| Dependency | Required? | Description |
|---|---|---|
| Python 3.8+ | ✅ Required | schema.py (frontmatter validation), store.py (SQLite data management), build_index.py (FTS5 indexing) are Python scripts. Uses only stdlib (sqlite3, json, pathlib) + pydantic |
| pydantic | ✅ Required | Frontmatter validation in schema.py. pip install pydantic |
| Filesystem write | ✅ Required | Creates and edits Markdown, SQLite, .obsidian/ config under .wiki/{topic}/. Will confirm location with user before first .wiki/ creation |
| WebSearch / WebFetch | ❌ Optional | Needed for active mode (agent searches for materials autonomously). Not needed for passive mode (user provides files) |
| External validator (MCP) | ❌ Optional | Only called during lint when wiki declares a validator. Silently skipped when unreachable, zero impact. No user credentials needed — Mcp-Session-Id is standard MCP protocol session handshake, handled automatically by the agent |
| Domain data MCP | ❌ Optional | deep-dive and active ingest can use domain data MCP for enhanced search quality. Falls back to WebSearch when absent |
Core promise: Passive mode (user provides files → agent compiles) only requires Python 3 + filesystem access, zero network dependencies. All network calls are optional enhancements, and the environment check on first use will inform the user of available capabilities.