Releases: ahmdd4vd/Fable5res
Release list
v1.2.0 — skills.sh Compatibility (YAML frontmatter)
Fable 5 Skills v1.2.0 — skills.sh Compatibility
The 5 Fable 5 skills are now installable via the open
skillsCLI (22k+ stars) and discoverable on skills.sh — the open agent skills directory by Vercel.
🆕 What's New
YAML frontmatter on all SKILL.md files
All 5 SKILL.md files now have proper YAML frontmatter (name + description), making them compatible with the Agent Skills specification used by the skills CLI:
---
name: fable-think
description: Think like Fable 5 — natural, flowing, purposeful reasoning...
---This enables automatic discovery by npx skills add ahmdd4vd/Fable5res --list.
Install via skills.sh
You can now install any of the 5 Fable 5 skills into 70+ supported coding agents using the skills CLI:
# List available skills
npx skills add ahmdd4vd/Fable5res --list
# Install a specific skill (e.g. fable-debug) into Claude Code + Cursor
npx skills add ahmdd4vd/Fable5res --skill fable-debug -a claude-code -a cursor
# Install all 5 skills into all detected agents
npx skills add ahmdd4vd/Fable5res --allThe skills CLI auto-detects which coding agents you have installed and installs the SKILL.md files into the correct paths for each.
Supported agents (70+)
Claude Code, Cursor, Cline, Codex, Windsurf, Continue, Gemini CLI, GitHub Copilot, OpenCode, Roo Code, Goose, Amp, Replit, Trae, Kimi Code CLI, Qwen Code, AiderDesk, AstrBot, Augment, CodeBuddy, Code Studio, Command Code, Cortex Code, Crush, Deep Agents, Devin, Droid, Firebender, ForgeCode, Hermes, inference.sh, Jazz, Junie, Kilo Code, Kiro CLI, Kode, Lingma, MCPJam, Mistral Vibe, Moxby, Mux, Neovate, Ona, OpenHands, Pi, Pochi, PromptScript, Qoder, Reasonix, Rovo Dev, Tabnine CLI, Terramind, Tinycloud, Zencoder, Zenflow, and more.
✅ Verified
Skills CLI discovery test
$ npx skills add ahmdd4vd/Fable5res --list
Source: https://github.com/ahmdd4vd/Fable5res.git
Repository cloned
Found 5 skills
Available Skills:
- fable-architect
- fable-code
- fable-debug
- fable-think
- fable-verify
Static skill verification (unchanged)
- Total claims verified: 31
- Passed: 31
- Failed: 0
- Pass rate: 100.00%
MiniMax M3 emulation (unchanged, Round 9)
- T1 reasoning: 100.00% (11/11) ✅
- T2 debug: 100.00% (9/9) ✅
- T3 code: 100.00% (8/8) ✅
- T4 verify: 100.00% (10/10) ✅
- T5 architect: 100.00% (8/8) ✅
- T6 self_correct: 100.00% (6/6) ✅
- T7 loop: 100.00% (7/7) ✅
- Average: 100.00% (Grade A)
📦 Install
Option 1: Original CLI (npm)
npx fable5-skills initOption 2: skills CLI (skills.sh)
npx skills add ahmdd4vd/Fable5res --all🔗 Links
- npm: https://www.npmjs.com/package/fable5-skills
- GitHub: https://github.com/ahmdd4vd/Fable5res
- skills.sh: https://skills.sh/ahmdd4vd/Fable5res (indexing in progress)
- Source dataset: https://huggingface.co/datasets/Kuberwastaken/Fable-5-traces
- Full changelog: https://github.com/ahmdd4vd/Fable5res/blob/main/CHANGELOG.md
v1.1.0 — AGPL-3.0 + README Redesign
Fable 5 Skills v1.1.0 — AGPL-3.0 + README Redesign
⚠️ Breaking license change: this release re-licensesfable5-skillsfrom MIT to AGPL-3.0-only to comply with the upstream dataset's license. See below for details.
🔄 License Change (breaking)
The source dataset Kuberwastaken/Fable-5-traces is licensed under AGPL-3.0. Because fable5-skills is derived from that dataset, this package inherits the AGPL-3.0 copyleft starting from v1.1.0.
| Aspect | v1.0.0 | v1.1.0 |
|---|---|---|
| License | MIT | AGPL-3.0-only |
| Network use clause | No | Yes (Section 13) |
| Enterprise-friendly | Yes | Requires dual-license discussion |
| Source disclosure on modification | No | Yes |
What this means for users
- ✅ You may use, study, modify, and redistribute this package.
- ✅ You may build network services on top of it.
⚠️ If you modify or redistribute (including as a network service), you must publish your source code under AGPL-3.0.- For enterprise use that cannot tolerate AGPL, contact the maintainer to discuss a dual-license arrangement.
🎨 README Redesign
Complete visual overhaul — modern, professional, elegant:
- Centered hero header with title, tagline, and navigation links.
- Modern badge row: npm version, downloads, AGPL license, Node engine, 31/31 verification, Grade A.
- Visual benchmark bars for all 7 MiniMax M3 test scenarios (ASCII progress bars).
- Iterative improvement curve table showing rounds 1–9 (80.82% → 100.00%).
- Collapsible detail sections for the behavioral profile:
- Chain-of-thought structure
- Linguistic signature
- Tool usage & verification
- Session-level patterns
- Clean tables with proper alignment for skills list, CLI flags, and benchmark results.
- Removed all third-party personal attributions per maintainer request.
✅ Verified (unchanged from v1.0.0)
Static skill verification
- Total claims verified: 31
- Passed: 31
- Failed: 0
- Pass rate: 100.00%
MiniMax M3 emulation (Round 9 — final)
| Test | Score | Status |
|---|---|---|
| T1 reasoning | 100.00% (11/11) | ✅ |
| T2 debug | 100.00% (9/9) | ✅ |
| T3 code | 100.00% (8/8) | ✅ |
| T4 verify | 100.00% (10/10) | ✅ |
| T5 architect | 100.00% (8/8) | ✅ |
| T6 self_correct | 100.00% (6/6) | ✅ |
| T7 loop | 100.00% (7/7) | ✅ |
| Average | 100.00% | ✅ Grade A |
📦 Install
npx fable5-skills initDefault installs into .claude/skills/ (Claude Code layout). Use --agent=cursor|cline|windsurf|continue|generic for other layouts.
🔗 Links
v1.0.0 — Grade A (100%) Release
Fable 5 Skills v1.0.0 — Grade A (100%) Release
AI reasoning skills distilled from 4,665 real Claude Fable 5 chain-of-thought traces, mathematically tuned and validated.
🎯 Highlights
- 5 skills (
fable-think,fable-code,fable-debug,fable-architect,fable-verify) — oneSKILL.mdeach, drop-in for any agent that reads markdown skills. - 31/31 quantified claims verified against the source dataset (100% pass rate).
- 9 iterative test rounds against MiniMax M3, final score 100.00% (Grade A).
- CLI installer (
npx fable5-skills init) drops the skills into the right directory for your agent in one command. - Multi-agent support: Claude Code, Cursor, Cline, Windsurf, Continue, or generic
./skills.
📦 Install
npx fable5-skills initDefault installs into .claude/skills/ (Claude Code layout). Use --agent=cursor|cline|windsurf|continue|generic for other layouts.
🧠 The 5 Skills
| Skill | Purpose | Core Loop |
|---|---|---|
fable-think |
Foundational per-turn reasoning | ACKNOWLEDGE → OBSERVE → EXECUTE → VERIFY |
fable-code |
Writing & editing code | Read → Understand → Plan → Write → Verify → Iterate |
fable-debug |
Root cause analysis | OBSERVE → INVESTIGATE → HYPOTHESIZE → ROOT CAUSE → FIX → VERIFY |
fable-architect |
System design | UNDERSTAND → DESIGN → VERTICAL SLICE → VERIFY → ITERATE |
fable-verify |
Quality assurance | 5-phrase verification vocabulary |
📊 Benchmark Results
Static skill verification (claims vs. actual data)
- Total claims verified: 31
- Passed: 31
- Failed: 0
- Pass rate: 100.00%
MiniMax M3 emulation (Round 9 — final)
| Test | Score | Status |
|---|---|---|
| T1 reasoning | 100.00% (11/11) | ✅ |
| T2 debug | 100.00% (9/9) | ✅ |
| T3 code | 100.00% (8/8) | ✅ |
| T4 verify | 100.00% (10/10) | ✅ |
| T5 architect | 100.00% (8/8) | ✅ |
| T6 self_correct | 100.00% (6/6) | ✅ |
| T7 loop | 100.00% (7/7) | ✅ |
| Average | 100.00% | ✅ Grade A |
🔬 Source Dataset
Distilled from Kuberwastaken/Fable-5-traces — 4,665 rows, 60 sessions of real Claude Fable 5 chain-of-thought traces.
🚀 Quick Start
# Install into Claude Code layout (.claude/skills/)
npx fable5-skills init
# Install into Cursor layout (.cursor/skills/)
npx fable5-skills init --agent=cursor
# Install a single skill
npx fable5-skills init --only=fable-debug
# Verify package integrity
npx fable5-skills doctor⚖️ License
MIT — see LICENSE.
🙏 Acknowledgements
- Anthropic PBC — for Claude Fable 5
Kuberwastaken/Fable-5-traces— source dataset- Matt Pocock — for the SKILL.md format
- MiniMax Inc. — for MiniMax M3, used as the cross-model emulation benchmark
npm: https://www.npmjs.com/package/fable5-skills
GitHub: https://github.com/ahmdd4vd/Fable5res