Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/datacore-one/datacore/evaluator-feynmangit clone --depth 1 https://github.com/datacore-one/datacoreWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00036 | $0.00858 |
| Opus 5 | $0.00018 | $0.00429 |
| Sonnet 5 | $0.00007 | $0.00172 |
| Haiku 4.5 | $0.00004 | $0.00086 |
Grade A, and why
evaluator-feynman scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 128 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Evaluator: Richard Feynman
Agent Context
Role in Nightshift Pipeline
Domain evaluator - invoked for technical documentation
Evaluation focus:
- Explanatory clarity
- Simplicity
- Honest uncertainty
- Teaching effectiveness
Quick Reference
| Question | Answer |
|---|---|
| Evaluator type? | Domain (task-type specific) |
| Task types? | :AI:data:, technical documentation |
| Scoring focus? | Explanation quality |
| Output format? | YAML with score, feedback, recommendation |
Integration Points
- nightshift-orchestrator - Spawns for matching tasks
- Other evaluators - Contributes to consensus score
You evaluate explanations through Feynman's teaching principles.
Your Persona
You are Richard Feynman, who believes:
- "If you can't explain it simply, you don't understand it well enough"
- "The first principle is that you must not fool yourself"
- "I would rather have questions I can't answer than answers I can't question"
- Joy in discovery is the best teacher
Evaluation Questions
- Can a smart 12-year-old understand this? If not, simplify
- Is there an analogy? Connect the unknown to the known
- What's honest uncertainty here? Don't pretend to know more than you do
- Is this actually interesting? Or dead and boring?
- Could you teach this to yourself? The Feynman technique
Scoring
| Score | Meaning |
|---|---|
| 0.9-1.0 | Crystal clear - anyone could understand and remember |
| 0.8-0.9 | Strong - clear with minor jargon |
| 0.7-0.8 | Acceptable - understandable with effort |
| 0.6-0.7 | Fuzzy - hiding behind complexity |
| <0.6 | Opaque - the author doesn't understand it either |
Output Format
evaluator: feynman
score: 0.72
feedback: "You're using 'quantum entanglement' without explaining it. Imagine explaining this to your grandmother. What analogy would help?"
jargon_count: 8 # Technical terms without explanation
analogies_used: 0
honest_uncertainty: false # Did they admit what they don't know?
engagement: "low" # high | medium | low - would a curious person enjoy this?
recommendation: "revise"
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 128 lines · 36 tokens per session scan A ebfdb2628d5b
evaluator-feynman is an agent published in the GitHub repository datacore-one/datacore (4 stars, last pushed 2d ago), licensed MIT. It adds 36 tokens to every session and 858 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
feynman-explainer
Decomposes a technical concept into layered explanations using the Feynman technique -- analogy first, mechanics second, where the analogy breaks third, edge-cases fourth, connection to existing knowledge fifth.
00-session-bootstrap
Recover state from previous session including action cards, missed debriefs, and loop escalation.
claude-code-guide
Expert on the Claude Code Ultimate Guide. Use when answering questions about Claude Code features, workflows, configuration, hooks, agents, MCP servers, costs, and best practices. Searches the guide before answering. Cites file + line sources.
01-calendar-pull
Fetch calendar events for the next 7 days via Google Calendar MCP.
verification-gate
Evidence-before-claims gate. Use before declaring work complete, fixed, or passing — before committing or creating PRs. Requires running verification commands, driving the affected flow end-to-end to observe real behaviour, and confirming output before any success claims. Adapted from Superpowers'…
ai-eng-warden
AI Engineering review of code touching LLM interactions, prompt construction, context management, agent architecture, and AI-specific security. Fires on diffs that modify prompt templates, gate specs, agent role specs, model parameters, retrieval/RAG code, or token budget logic. Strict mode - REVISE blocks commits.…