xr843/Master-skill
FoJin-powered Buddhist AI persona framework — source-grounded, boundary-aware, fidelity-tested, runtime-ready.
Master‑skill – AI‑driven Buddhist master personas
What it is
- An open‑source AgentSkill package that lets large‑language‑model assistants (Claude Code, Cursor, Codex, OpenCode, Gemini CLI, etc.) adopt the voice of historic Buddhist masters. Each “master” is a verified persona that answers questions by pulling directly from canonical Buddhist texts (CBETA, BDRC, SuttaCentral, PTS, etc.) and always cites the source.
- The framework is built on the FoJin text‑aggregation platform, which supplies >500 data sources, 678 k+ semantic vectors and a knowledge‑graph for real‑time retrieval.
- It follows the Anthropic AgentSkills standard, exposing the personas as slash‑commands (e.g.
/master-huineng) that can be installed with a singlenpxcall.
Key features
| Feature | Why it matters |
|---|---|
| 15 pre‑packaged masters spanning Indian, Chinese, Tibetan and Theravāda traditions | Users can instantly talk to a historically accurate figure such as Nagarjuna, Huineng, Milarepa, or Ajahn Chah. |
| Source‑grounded answers – every doctrinal claim is attached to a CBETA, BDRC, Toh, SuttaCentral or PTS identifier and a live FoJin link | Guarantees traceability and avoids hallucinated quotations. |
| Hard‑gate ethics layer – built‑in rules that block fabricating sources, giving esoteric instructions, or making personal attainment judgments | Keeps the system within clearly defined religious and copyright boundaries. |
| Two‑stage fidelity testing – JSON‑L test suites for each master, CI that runs dry‑run checks and optional Anthropic‑API scoring | Provides an objective measure of how well the persona stays faithful to its source material. |
Cross‑tradition tools – /compare‑masters, /master‑debate, /master‑curriculum |
Enables side‑by‑side doctrinal comparison, simulated debates, and curated study paths. |
| Multi‑platform deployment – pre‑built AgentSkill bundles for Claude Code, Cursor, Codex, OpenCode, Gemini CLI; also a Rust‑based desktop manager | Users can run the skills in a terminal, a web chat, or a native GUI. |
NPX one‑click install – npx master-skill install master-huineng (or --all) |
Removes any need for manual setup for most end‑users. |
Offline fallback – sources/ directory contains essential text excerpts; scripts/cite.py and scripts/query.py allow local citation and semantic search when FoJin is unavailable |
Guarantees functionality without network access. |
How to get started (developer install)
# Install all available master skills globally for Claude Code / Cursor etc.
npx master-skill install --all
# Or install a single master, e.g. Huineng
npx master-skill install master-huineng
# List what is installed
npx master-skill list
After installation, invoke the skill inside a supported LLM session with a slash command, e.g. /master-huineng "How should I deal with a restless mind?".
Typical use cases
- Buddhist study – learners can ask a specific master for guidance on meditation, doctrine, or scriptural references and receive answers that include exact canonical citations.
- Comparative research – scholars can run
/compare‑mastersto see how Indian Madhyamaka, Chinese Tiantai, Tibetan Gelug, and Theravāda perspectives differ on a topic. - Curriculum planning – the
/master‑curriculumcommand suggests a step‑by‑step study path across traditions, complete with source references. - Application integration – developers can embed the pre‑built AgentSkill bundles into their own AI assistants, ensuring the assistant respects religious and copyright boundaries.
Limitations & caveats
- All outputs are AI‑generated; they are not actual historic utterances and should not replace guidance from qualified teachers.
- The fidelity test suite currently covers ~70 % of the provided Q&A fixtures; many edge cases remain untested.
- Access to the full FoJin knowledge base requires an API key; offline mode falls back to a limited excerpt set.
- Ethical and copyright constraints are enforced by hard‑coded rules, but contributors must still follow the detailed
ETHICS.mdprocess when adding new masters.
Where to learn more
- Full documentation:
docs/(architecture, installation, roadmap) - Source‑grounding details:
ETHICS.mdandtests/fidelity.jsonl - Live demo (no install needed): https://fojin.app/chat – select “法师模式” and pick a master.
Master‑skill bridges large‑language‑model agents with rigorously sourced Buddhist scholarship, offering a reproducible, ethically‑guarded persona framework for both casual learners and academic researchers.
Related
- Project
- Project
- Project
- Dispatch
- Project