soulforge
Run high-signal autonomous coding loops with Soulforge (feature-dev/bugfix/review-loop) using strict worktree isolation, review gates, and scoped fix cycles.
- License 许可
- MIT
- 最近更新
- 2026-06-07
- Stars
- ★ 0
- Forks
- 1
01 前言 Front Matter
- name
- soulforge
- metadata
- { "openclaw": { "emoji": "🔥", "requires": { "env": [], "bins": [ "soulforge", "codex", "gh" ] } } }
- description
- Run high-signal autonomous coding loops with Soulforge (feature-dev/bugfix/review-loop) using strict worktree isolation, review gates, and scoped fix cycles.
03 SKILL.md 正文
Soulforge (Effective Use Guide)
This is not a full engine reference. This is the operating playbook for getting high-quality autonomous coding outcomes with Soulforge.
Core Operating Model
Use Soulforge when you want: plan → implement → verify/test → PR → review/fix loops with minimal babysitting.
Preferred workflows:
feature-devfor end-to-end feature deliverybugfixfor diagnose-first, surgical fixesreview-loopfor tightening an existing PR until clean
Golden Rules (Most Important)
- Never run from repo main checkout.
- Soulforge now enforces this guardrail.
- Always isolate work in worktrees.
- Default base:
<repo>/worktrees/
- Default base:
- Keep tasks tightly scoped.
- Specific issue, explicit acceptance criteria, explicit DO-NOT list.
- Treat review findings with discipline.
- FIX in-scope issues.
- Mark genuine extras as
SEPARATE.
- Use callback-exec only.
- HTTP callback mode is removed.
Current Behavior You Should Rely On
Workdir / Worktree safety
- If
--workdiris omitted, Soulforge can auto-provision a worktree under<repo>/worktrees/.... - Main checkout is blocked (including bare+worktree edge cases).
- Dirty worktrees are rejected for run start.
- Out-of-base workdirs are blocked unless explicitly overridden.
Checkpoint model
approve/rejectis gone.- Use structured completion via
soulforge complete .... - Pause checkpoints are
type: pause.
Callback model
- Use
--callback-exec. - Template vars include:
{{run_id}},{{step_id}},{{step_status}},{{status}},{{task}}{{callback_message}}(step-level, preferred){{prompt}}remains for backward compatibility in pause scenarios
Recommended Command Patterns
Feature build
soulforge run feature-dev "Implement <issue-url>.
Constraints: max 2 stories. DO NOT refactor unrelated modules." \
--workdir /abs/path/to/repo/worktrees/feat-xyz \
--callback-exec 'openclaw agent --session-key "agent:cpto:slack:channel:c0af7b05h28" --message "Soulforge {{run_id}} {{step_id}} {{step_status}}" --deliver'
Bugfix
soulforge run bugfix "Fix <issue-url> with failing test first; minimal patch only." \
--workdir /abs/path/to/repo/worktrees/fix-xyz \
--callback-exec 'openclaw agent --session-key "agent:cpto:slack:channel:c0af7b05h28" --message "Soulforge {{run_id}} {{step_id}} {{step_status}}" --deliver'
Review-only tightening on an existing PR
soulforge run review-loop "Review PR #123 and fix only in-scope findings." \
--workdir /abs/path/to/repo/worktrees/pr-123 \
--var pr_number=123 \
--callback-exec 'openclaw agent --session-key "agent:cpto:slack:channel:c0af7b05h28" --message "Soulforge {{run_id}} {{step_id}} {{step_status}}" --deliver'
How to Maximize Autonomous Quality
1) Give a tight task contract
Include:
- target issue/PR URL
- explicit in-scope list
- explicit out-of-scope list
- objective success criteria
2) Keep iteration loops short
If a PR loops repeatedly:
- create/update
.soulforge-progress.mdin worktree with exact outstanding fixes - run
review-loopconstrained to remaining findings
3) Handle gates like an operator, not a coder
At review gate:
- move in-scope defects to FIX
- separate unrelated ideas into follow-up issues
- avoid “while we’re here” drift
4) Expect long fix steps; optimize signal
Long fix steps are normal for real refactors. Your job is quality control at gates, not interrupting active runs.
Practical Triage Heuristic
When code-review returns findings:
- High/Medium tied to original issue: FIX now
- Low tied to original issue correctness: usually FIX now
- Anything outside scope: SEPARATE
Anti-Patterns (Avoid)
- Running multiple workflows in the same checkout
- Allowing scope creep in repeated review-fix loops
- Merging with known Highs because “tests pass”
- Treating this skill as generic Soulforge docs instead of an execution playbook
Minimal Status Workflow for Operator
- Start run
- Wait for review gate
- Triage with strict scope discipline
- Repeat until pass
- Merge
- Pull main + build + npm link + daemon restart (when local runtime should track latest)
Notes
- If loops hit
max_loops, spawn a fresh constrainedreview-looprun with a scope lock file. - For long-running initiatives, keep a brief run ledger in the channel (run id → PR → status).
04 同仓库更多 Skills
crowd-prompting
A marketplace where AI agents improve prompts, system instructions, tool descriptions, and other text-based content with domain expertise from real-world operations — and earn tokens for valuable contributions.
ceaser
Interact with the Ceaser privacy protocol on Base L2 using the ceaser-mcp MCP tools. This skill uses the ceaser-mcp npm package for ALL operations -- shield, unshield, note management, and protocol queries. All ceaser tool calls use CLI subcommands (npx -y ceaser-mcp <subcommand>). Ceaser is a privacy-preserving ETH wrapper using Noir/UltraHonk zero-knowledge proofs on Base L2 (chain ID 8453).
ceaser-send
Fully automated private ETH transfer via Ceaser Protocol on Base L2 using the ceaser-mcp MCP tools. This skill uses the ceaser-mcp npm package for ALL operations -- all ceaser tool calls use CLI subcommands (npx -y ceaser-mcp <subcommand>). WARNING -- creates observable on-chain links between user wallet and hot wallet (see Privacy Warning). Generates an ephemeral hot wallet with BIP-39 mnemonic recovery, user funds it, agent signs and broadcasts the Shield TX automatically. Exactly one manual step required (funding the hot wallet with ETH). Uses Noir/UltraHonk zero-knowledge proofs.
dilemm-ai
Compete in DilemmAI, the prisoner's dilemma AI arena at dilemm.ai. Use when an OpenClaw agent wants to create an account, design and submit strategy prompts for their bot, enter matchmaking, analyze historical matches and leaderboards, and iterate on bot strategy. Requires headless browser automation (no API available yet).
moldium
Post and manage content on the Moldium blog platform. Triggered by "post to Moldium", "write a blog post", "publish an article", etc.
arxiv-paper-reviews
Interact with arXiv Crawler API to fetch papers, read reviews, submit comments, search papers, and import papers. Use when working with arXiv papers, fetching paper lists by date/category/interest, viewing paper details with comments, submitting paper reviews, searching papers by title, or importing papers from arXiv URLs via API at http://weakaccept.top:8000/.