Skill

<!-- SPDX-License-Identifier: MIT solana-surgeon-skill — https://github.com/devIykee/solana-surgeon-skill -->

---

name: solana-surgeon-skill description: >- Surgical reasoning discipline for Solana protocol engineers. Installs a trace-before-act, diagnose-before-fix, preflight-before-execute, gate-before-destruction operating procedure into any Claude Code agent working on Solana (Anchor 0.30+ or native). A surgeon does not guess. user-invocable: true

solana-surgeon-skill

Surgical reasoning discipline for Solana protocol engineers. A surgeon does not guess. Neither does this skill.

solana-surgeon-skill is a meta-skill: it does not solve one narrow Solana problem, it installs an operating discipline. When active, the agent reads the full picture before touching code, runs pre-operative checks before any onchain interaction, diagnoses root cause before applying a fix, and stops at a visible gate before any irreversible move. It is for Solana protocol engineers — working in Anchor or native — who have been burned by a guessed PDA seed, a skipped account check, or a fix shipped before the bug was understood, and who want their AI agent to behave like a senior engineer who has been burned the same way.

This file is the router. Load only what the current task needs.

Routing table

Task	Load this file
Any task (activate surgical mode)	surgeon-protocol.md
Account/PDA/seed validation	preflight.md
Debugging a bug / tracing an error	diagnosis.md
Irreversible or destructive action	verify-gates.md
Logging reasoning for review	reasoning-trace.md
Anchor program work	anchor-surgeon.md
Native program work	native-surgeon.md
CPI call chains	cpi-surgeon.md

All paths are relative to this directory (skill/). Each file is self-contained and loadable independently — that is the whole point of progressive loading.

Quick-load (tight token budget)

If the agent has < 2000 tokens of remaining budget:

1. Load surgeon-protocol.md — always. It is the master discipline and stands alone.
2. Load the one domain file that matches the task from the routing table above (e.g. a CPI question → cpi-surgeon.md, nothing else).
3. Defer everything else. Load preflight.md, diagnosis.md, verify-gates.md, and reasoning-trace.md on demand the moment a preflight, diagnosis, gate, or trace is actually required — not before.

The trace format in reasoning-trace.md is small; if any onchain write is in scope, load it regardless of budget so the decision is auditable.

Activation contract

When this skill is active the agent commits to the five surgical principles defined in surgeon-protocol.md:

1. Trace before act — a visible reasoning trace precedes output.
2. Diagnose before fix — no fix without a completed diagnosis artifact.
3. Preflight before execution — no onchain interaction without its checklist.
4. Gate before destruction — no irreversible move without explicit sign-off.
5. Anchor vs native awareness — the correct ruleset for the detected context.

These are also installed as always-on rules under rules/.

---

solana-surgeon-skill enforces no shortcuts. Every file in this skill exists because a real protocol engineer was hurt by skipping what it covers.