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 skills add supportersimulator/3-surgeons --skill using-3-surgeonsgit clone --depth 1 https://github.com/supportersimulator/3-surgeonsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/supportersimulator/3-surgeons/using-3-surgeons)<a href="https://agentmods.dev/skills/supportersimulator/3-surgeons/using-3-surgeons"><img src="https://agentmods.dev/badge/skills/supportersimulator/3-surgeons/using-3-surgeons.svg" alt="Measured on agentmods" height="20"></a>What 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 5.1 | $0.00034 | $0.03737 |
| Opus 5 | $0.00017 | $0.01869 |
| Sonnet 5 | $0.00007 | $0.00747 |
| Haiku 4.5 | $0.00003 | $0.00374 |
Grade A, and why
using-3-surgeons 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 8d 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 — 313 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Using 3-Surgeons
Philosophy
"The value is in the disagreements, not the agreements."
Three LLMs with different training, different biases, and different blind spots cross-examine each other on critical decisions. Claude may confabulate confidently where GPT hedges. A local 4B model catches patterns that API models miss due to token limits. When all three agree, confidence is high. When they disagree, you have found the exact spot that needs human judgment.
The Three Surgeons
| Surgeon | Model | Role | Cost |
|---|---|---|---|
| Atlas (Head Surgeon) | Claude (this session) | Synthesizes, decides, implements | $0 (session) |
| Cardiologist | GPT-4.1-mini (OpenAI API) | External perspective, cross-examination, evidence review | ~$0.40-1.60/1M tokens |
| Neurologist | Qwen3-4B (local Ollama/MLX) | Pattern recognition, corrigibility checks, classification | $0 (local) |
Atlas is always available -- it is the Claude session itself. The Cardiologist and Neurologist are external models accessed via the 3s CLI or MCP tools.
Constitutional Physics
These principles outrank all other preferences:
- Preserve Determinism -- same inputs must produce the same outputs. If not, enter safe mode.
- No Discovery at Injection -- injection is retrieval + assembly only. No learning or promotion during injection.
- Evidence Over Confidence -- outcomes determine truth. Confidence tracks empirical rates, not feelings.
- Prefer Reversible Actions -- checkpoint before risk, canary before rollout.
- Minimalism -- maximum value density. Scalpel, not axe.
Tool Access
All 3-Surgeons operations are available through two interfaces:
CLI (3s command)
3s probe # Health check all surgeons
3s cross-exam "topic" # Full cross-examination (uses config default mode)
3s cross-exam "topic" --mode iterative # Cross-exam with explicit review depth
3s consult "topic" # Quick parallel query
3s consensus "claim" # Confidence-weighted vote
3s sentinel "content" # Complexity vector scan
3s gains-gate # Infrastructure health gate
3s ab-propose PARAM A B "hyp" # Propose A/B test
3s mode # Show current review depth + auto-depth setting
3s mode continuous # Set default review depth
3s mode iterative --duration 7d # Set mode with expiry (session|7d|30d|permanent)
3s review-weights # Show learned mode weights
3s review-weights show # Same as above
3s review-weights export -o weights.json # Export outcomes for sharing
3s review-weights import weights.json # Import outcomes from another machine
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.
- 8d ago First seen · 313 lines · 34 tokens per session scan A bf90e3275a10
using-3-surgeons is a skill published in the GitHub repository supportersimulator/3-surgeons (2 stars, last pushed 3mo ago), licensed MIT. It adds 34 tokens to every session and 3,737 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 skills, from other repositories
gh-pr-review
Automated Cherry Studio review for local branches, PRs, commits, files, architecture docs, and repository skills. Use for code or documentation reviews that need project-specific naming, main/renderer/shared placement and dependency rules, IpcApi and DataApi boundaries, lifecycle/service ownership, renderer hooks…
review-spd
Findings-first code review workflow for AI coding agents. Use when the user asks to review uncommitted changes, commits in a date range, or a branch compared to the main branch / PR-style diff. Focuses on bugs, regressions, correctness risks, missing tests, security/data-safety issues, and other behavior-changing…
karpathy-guidelines
A set of coding guidelines based on observations about common mistakes made by language models. It emphasizes simple solutions, small targeted edits, clear assumptions, and checkable results.
gh-cli
Patterns for invoking the GitHub CLI (gh v2.100.0+) from agents. Use when the task mentions GitHub, gh, pull requests/PRs, issues, releases, gists, Actions/workflow runs, forks, repo cloning, reviews, or you need exact gh commands. Covers pagination, repo targeting, search vs list, discussions, projects, rulesets…
code-review
A single-pass review method for examining a code change, branch, or pull request. A pull request is a proposed change for review before it is merged into a shared codebase.
simplify
Behavior-preserving code simplification — reduce complexity without changing what the code does. Use when the task mentions "simplify", "reduce complexity", "too clever", "hard to read", "reduce nesting", or after a feature lands and the code needs polishing. Oracle analyzes (read-only), light-orchestrator applies the…