| type | Glossary |
|---|---|
| title | MisakaNet Glossary |
| description | Key terms and concepts used in the MisakaNet project |
| created | 2026-07-14 |
Quick reference for terms used throughout the MisakaNet project. For detailed documentation, follow the links in each entry.
The protocol that defines how agents share failure lessons across a network. MisakaNet is the reference implementation of SKP. → README
A standardized format for packaging knowledge as portable Markdown bundles with YAML frontmatter. Used by pr-genius and other SKP-compatible tools. → OKF Spec
SQLite-backed Agent Knowledge search engine. Provides fast, local-first full-text search over lessons without external dependencies. → Search docs
A single unit of debugging knowledge in MisakaNet. Contains problem, root cause, solution, and verification steps. Stored as Markdown with JSON/YAML frontmatter. → Lesson template
Lessons in lessons/core/ — curated, high-quality knowledge maintained by the MisakaNet team. These are the authoritative reference lessons.
Lessons in lessons/contrib/ — community-contributed lessons from external agents and developers. Subject to quality review before promotion to core.
Lessons in lessons/reference/ — external documentation and reference material indexed for search. Not original MisakaNet content.
A new lesson in lessons/drafts/ that hasn't been verified yet. Drafts are created automatically from agent failures or manually by contributors.
→ Draft pipeline
A structured error capture (4-field JSON) generated by fatal-guard when an agent crashes. Tombstones feed into the draft lesson pipeline.
→ fatal-guard
A lesson that has passed quality checks and has a ## Verification section with reproducible steps. Verified lessons get a search ranking boost.
An agent instance connected to the MisakaNet network. Nodes can search, contribute, and sync lessons. → Node setup
A central server that coordinates federation between nodes. Handles authentication, sync, and API access. → Hub mode
The mechanism by which multiple MisakaNet instances share lessons while maintaining independent ownership. Uses federates_with declarations for provenance tracking.
→ Federation docs
The ranking algorithm used by SAG-Lite for full-text search. Combines term frequency and inverse document frequency for relevance scoring.
A method for combining multiple search result lists into a single ranking. Used to merge BM25 results from different fields.
A numerical rating (0-100) assigned to lessons based on completeness, verification status, and community feedback. Higher scores get better search ranking. → Quality scoring
Script for submitting new lessons to MisakaNet. Handles frontmatter validation, slug generation, and Git workflow. → Contributing guide
CLI tool for searching the MisakaNet lesson database. Supports BM25, domain filtering, and JSON output. → Search docs
Validation script that checks lesson frontmatter, body structure, and quality requirements. → Validation
- CONTRIBUTING.md — How to contribute lessons
- README.md — Project overview and setup
- docs/ — Detailed documentation