by proxysoul
AI coding agent that edits symbols, not strings. AST surgery, full LSP, and a live code graph wired to memory that resurfaces by file, co-change, and semantics.
# Add to your Claude Code skills
git clone https://github.com/proxysoul/soulforgeLast scanned: 5/10/2026
{
"issues": [
{
"type": "npm-audit",
"message": "@jimp/core: Vulnerability found",
"severity": "medium"
},
{
"type": "npm-audit",
"message": "@jimp/diff: Vulnerability found",
"severity": "medium"
},
{
"type": "npm-audit",
"message": "@jimp/js-bmp: Vulnerability found",
"severity": "medium"
},
{
"type": "npm-audit",
"message": "@jimp/js-gif: Vulnerability found",
"severity": "medium"
},
{
"type": "npm-audit",
"message": "@jimp/js-jpeg: Vulnerability found",
"severity": "medium"
},
{
"type": "npm-audit",
"message": "@jimp/js-png: Vulnerability found",
"severity": "medium"
},
{
"type": "npm-audit",
"message": "@jimp/js-tiff: Vulnerability found",
"severity": "medium"
},
{
"type": "npm-audit",
"message": "@jimp/plugin-blur: Vulnerability found",
"severity": "medium"
},
{
"type": "npm-audit",
"message": "@jimp/plugin-color: Vulnerability found",
"severity": "medium"
},
{
"type": "npm-audit",
"message": "@jimp/plugin-contain: Vulnerability found",
"severity": "medium"
},
{
"type": "npm-audit",
"message": "@jimp/plugin-cover: Vulnerability found",
"severity": "medium"
},
{
"type": "npm-audit",
"message": "@jimp/plugin-crop: Vulnerability found",
"severity": "medium"
},
{
"type": "npm-audit",
"message": "@jimp/plugin-hash: Vulnerability found",
"severity": "medium"
},
{
"type": "npm-audit",
"message": "@jimp/plugin-print: Vulnerability found",
"severity": "medium"
},
{
"type": "npm-audit",
"message": "@jimp/plugin-resize: Vulnerability found",
"severity": "medium"
},
{
"type": "npm-audit",
"message": "@jimp/plugin-rotate: Vulnerability found",
"severity": "medium"
},
{
"type": "npm-audit",
"message": "@jimp/plugin-threshold: Vulnerability found",
"severity": "medium"
},
{
"type": "npm-audit",
"message": "@opentui/core: Vulnerability found",
"severity": "medium"
},
{
"type": "npm-audit",
"message": "@opentui/react: Vulnerability found",
"severity": "low"
},
{
"type": "npm-audit",
"message": "diff: jsdiff has a Denial of Service vulnerability in parsePatch and applyPatch",
"severity": "low"
},
{
"type": "npm-audit",
"message": "file-type: file-type affected by infinite loop in ASF parser on malformed input with zero-size sub-header",
"severity": "medium"
},
{
"type": "npm-audit",
"message": "jimp: Vulnerability found",
"severity": "medium"
}
],
"status": "PASSED",
"scannedAt": "2026-05-10T06:36:59.099Z",
"semgrepRan": false,
"npmAuditRan": true,
"pipAuditRan": true
}soulforge is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by proxysoul. AI coding agent that edits symbols, not strings. AST surgery, full LSP, and a live code graph wired to memory that resurfaces by file, co-change, and semantics. It has 831 GitHub stars.
Yes. soulforge passed SkillsLLM's automated security scan — a dependency vulnerability audit plus prompt-injection heuristics — with no high-severity issues. You can read the full report in the Security Report section on this page.
Clone the repository with "git clone https://github.com/proxysoul/soulforge" and add it to your Claude Code skills directory (see the Installation section above).
soulforge is primarily written in TypeScript. It is open-source under proxysoul on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh soulforge against similar tools.
No comments yet. Be the first to share your thoughts!
brew tap proxysoul/tap && brew install soulforge
# alternatives
bun install -g @proxysoul/soulforge
# or download a prebuilt binary from https://github.com/ProxySoul/soulforge/releases/latest
macOS, Linux, and Windows 10 1809+ / Windows 11 (x64). On Windows, grab soulforge-<version>-windows-x64.zip (portable) or soulforge-setup-<version>-x64.exe (setup wizard) from the latest release. ARM64 not yet supported (tracked upstream).
soulforge --set-key anthropic sk-ant-...
cd your-project
soulforge
Other providers and OpenAI-compatible endpoints: docs/providers.
Same model (Claude Opus 4.6), same codebase, same prompt.
Bug fix
| SoulForge | OpenCode | |
|---|---|---|
| Time | 6m 22s | 11m 18s |
| Cost | $1.70 | $3.52 |
| Result | Correct | Correct |
Audit task ("verify cost reporting is wired correctly")
| SoulForge | OpenCode | |
|---|---|---|
| Time | 2m 00s | 5m 56s |
| Cost | $0.84 | $2.61 |
| Accuracy | 7/7 (100%) | 4/7 (57%) |
| False alarms | 0 | 3 |
| Wrong claims | 0 | 1 |
Same bug. Same model. Same repo. Half the time. Half the cost.
Sources: recording 1 · recording 2 · recording 3
| Feature | What it does |
|---|---|
| AST editing | TS/JS edits via ts-morph, 65+ ops, atomic batches. docs |
| Live Soul Map | SQLite graph, PageRank + git co-change, blast-radius tags. docs |
| LSP + Mason | 576+ servers installable from the TUI |
| 33 languages | symbol-level reads, not file dumps |
| Compound tools | rename_symbol, move_symbol, refactor, project (23 toolchains) |
| Task router | route each slot (spark / ember / compact / verify / web / semantic) to a different model per tab. Haiku for exploration, Sonnet for code, Flash for compaction. /router. docs |
| V2 compaction | usually 0 LLM tokens. docs |
| Parallel agents | Spark + Ember with shared I/O cache |
| Embedded Neovim | real nvim in a PTY, your config |
| 5 tabs | per-tab model, session, checkpoints, file claims |
| Time machine | every prompt is a checkpoint with a git tag. Ctrl+B / Ctrl+F rewinds and redoes both conversation and files on disk. /checkpoint undo <N>, /checkpoint save, per-tab |
| Sessions | auto-saved JSONL, crash-resilient, resumable by short-id prefix. Export to markdown / JSON / clipboard. Ctrl+P browser, multi-tab. docs |
| Memory | cross-session SQLite store of prefs, decisions, gotchas, context. Auto-recalled per turn from prompt + edited files. Project + global scopes, browser at /memory. docs |
| 21 providers | Anthropic, OpenAI, Google, Groq, DeepSeek, Bedrock, Ollama, LM Studio, ... + any OpenAI-compatible |
| Cost tracking | per-model + per-subagent USD, cache-aware |
| MCP + hooks | any MCP server, 13 events, drop-in .claude/settings.json |
| Headless mode | run from CI, scripts, pipelines. JSON / event stream, resumable sessions. docs |
| Hearth (exp) | remote control via Telegram or Discord, your host only |
| 36 themes | hot-reloaded JSON, Kitty inline images |
BSL 1.1. Free for personal and internal use. Commercial use: commercial license. Converts to Apache 2.0 on March 15, 2030.
Backed by
Sponsor on GitHub (monthly or one-time) · PayPal (one-time) · All backers