Traction scoreGitHub stars can be faked, so popularity alone can be misleading. Traction Score looks for broader signs of recent attention, adoption, and active maintenance.
Fable Mode provides a structured discipline for AI agents tackling large, open-ended tasks. It replaces immediate execution with a requirement for written stage plans and explicit delegation to specialized agents, ensuring that complex workflows are broken down into verifiable steps.
This approach is designed for automation builders who need to reduce fabrication in research and enforce rigorous verification guardrails at lower execution tiers.
WHO IT'S FOR
AI agent / automation builders
implementing multi-stage planning and delegation workflows
Technical leads / architects
establishing rigorous AI execution guardrails and verification
AI power users
executing complex, multi-file research and coding tasks
Platform / DevEx teams
optimizing LLM performance across different model tiers
Compatible AgentsThe repository documents support for these agents. The skills may also work with other agents that can load SKILL.md files, but they may need some setup or small changes.
Follow the documented setup, then try a first task.
I’m using
Claude Code user-scope bundle
Installs: Complete fable-mode skill family and its agent definitions, installed at user scope under ~/.claude/skills/ and ~/.claude/agents/. · Claude Code
Before you start
Git must be available to run the documented clone command.
A Bash-compatible environment must be available to execute install.sh.
Platform: Claude Code
In a terminal, run the documented command from the directory where you want the fable-mode repository cloned. It clones the repository, enters it, and runs the user-scope installer.
git clone https://github.com/mrtooher/fable-mode.git && cd fable-mode && ./install.sh
Installs: Complete seven-skill fable-mode family installed individually through Settings; agent definitions are excluded, so the skills use their documented inline fallback. · Cowork, claude.ai
Before you start
Access to the repository's SKILL.md files.
Platform: Cowork / claude.ai
Open Settings → Capabilities → Skills and add each skill file individually. Replace the documented <name> placeholder with fable-fable, fable-opus, fable-sonnet, fable-haiku, execution-guardrails, and double-check; add the repository-root SKILL.md as fable-mode. Keep each installed folder name equal to its SKILL.md name field. Do not add files from agents/ as skills; those agents are unavailable on this surface and the skills fall back to inline mode.
Two source files replacing [source path A] and [source path B]
A destination replacing [output path]
The comparison question or decision the brief should address
Use the fable-mode skill to produce a short brief comparing [source path A] and [source path B]. Write the brief to [output path]. Start with a numbered stage map, give each stage a check that can fail, map every load-bearing claim to one of the supplied sources, perform the required self-critique, and apply the mandatory double-check delivery gate before presenting the result.
Its requirements or acceptance criteria replacing [spec path]
Use the double-check skill to verify the finished deliverable at [artifact path] against the requirements in [spec path]. Check the existing output rather than redoing the task. Fix any confirmed FAIL and re-run its specific failed check, then provide the verification summary. Run only one double-check on this deliverable.
Use the execution-guardrails skill while replacing the exact word [old word] with [new word] in [file path]. First verify that the target occurrence exists. Use a unique targeted anchor when possible; otherwise use word-boundary matching, never a bare substring replacement. Afterward, check the file for glued or malformed compound words and report only confirmed problems.
Claude