Most code simplification tools check for style violations and linting. Simplify Swarm breaks your code down into risk tiers and cleans with surgical precision.
Three parallel subagents attack your diff from different angles: Hygiene hunts dead code, AI slop, pass-through wrappers, and stale state — the stuff that rots silently. Clarity goes after duplication, naming rot, nested ternaries, and structural bloat — the stuff that makes your teammates sigh. Correctness traces N+1 queries, memory leaks, concurrency races, leaky abstractions, and silent failures — the stuff that breaks at 3am.
Then a consolidator merges, de-duplicates, and triages everything into a risk-tiered application order: SAFE (auto-applied — unused imports, dead branches), CAREFUL (applied with test verification — renames, extracted helpers), RISKY (flagged for human review — N+1 restructures, public API changes, concurrency fixes).
It's read-only during analysis — finds problems, doesn't touch your code until you've seen the consolidated report. Opt-in only, no auto-trigger. Run /simplify when you want to know what you actually shipped.
Built by synthesising 11 implementations: Anthropic's official code-simplifier plugin, nicknisi's essentials fork (AI slop detection, dead code removal), Navigator's ROI-gated config-driven approach, Addy Osmani's language-specific guidance and Chesterton's Fence, OMC's ai-slop-cleaner and 4-pass workflow, ECC's refactor-cleaner (knip/depcheck/ts-prune), WalletConnect's early-return patterns, Meta OpenEnv's TDD Red-Green-Refactor integration, GitHub Copilot's structured output format, and the Claude Code /simplify stop-hook architecture.
Langauge-specific detectors for TypeScript/JavaScript, Python, and Go. N+1 query patterns. Memory leak signatures. Concurrency anti-patterns. Silent failure spots. All the things a general-purpose reviewer misses because it's looking too broad.
Companion to Hermes Agent's built-in simplify-code skill. That one's 175 lines, general-purpose, fast. This one's the deep clean — use it before PRs, before deploys, when you inherit code you don't trust.
Upstream enhancement PR is open with the 5 highest-value additions (risk-tiered application, Chesterton's Fence, AI slop detection, silent failure detection, structured output): https://github.com/NousResearch/hermes-agent/compare/main...Sahil-SS9/KenseiAgent:feat/enhance-simplify-code
If you write a skill that pairs with this (CI integration, auto-test generator for findings, language pack for Rust/Zig/Kotlin), open a PR. Bug reports appreciated. If the Correctness agent flagged your code and you're annoyed — it's designed to be thorough.
mkdir -p ~/.hermes/skills/software-development
cd ~/.hermes/skills/software-development
git clone https://github.com/Sahil-SS9/hermes-simplify-swarm.git simplify-swarmNo dependencies. No config required. Works on any Hermes Agent install.
"simplify my changes"
/cleanup
"deslop this module"
"/simplify focus on correctness"
The skill auto-detects scope via git diff. SAFE changes applied automatically. CAREFUL changes applied with test verification. RISKY findings flagged for you to decide.
| Agent | What It Hunts | Examples |
|---|---|---|
| Hygiene (SAFE) | Dead code, AI slop, redundant abstractions, stale state, utility discovery | Unused imports, as any casts, pass-through wrappers, duplicate state stores |
| Clarity (CAREFUL) | Duplication, naming, structure, consistency | Nested ternaries, data/temp variables, 80-line functions, mixed arrow/function styles |
| Correctness (RISKY) | N+1 queries, memory leaks, concurrency, leaky abstractions, silent failures | ORM queries in loops, unclosed resources, race conditions, empty catch blocks |
Optional .kensei/simplify.yaml in your project root:
simplify:
auto_apply: true # Skip approval for SAFE+CAREFUL tiers
skip_patterns:
- "*.test.*"
- "*.generated.*"
max_file_lines: 800Built by Sahil Saghir (KENSEI). Inspired by Claude Code's /simplify and the upstream Hermes Agent simplify-code skill.
MIT — use it, fork it, ship it.