Run trap-aware agent work through the Fable Method
Use the Fable Method skill bundle to classify work, gather evidence, act narrowly, verify by observation, and judge agent outputs against repeatable eval-backed rules.
npx skills add agentskillexchange/skills --skill run-trap-aware-agent-work-through-the-fable-method
/plugin install fable@fable-method or clone the repository and run bash fable-method/install.sh. For Codex, Cursor, aider, or raw agent prompts, use the repository's AGENTS.md and bundled skills as the invocation instructions.Use the Fable Method when an agent must handle work where false completion, ambiguous authority, skipped verification, or weak execution would be expensive. The workflow is concrete: apply the fable-method loop to classify the ask, define done, gather primary evidence, make one recommendation, act narrowly, verify with an observed check, and report caveats; use fable-loop for unattended or delegated work, fable-judge after an agent claims completion, and fable-domain to create a domain adapter when the recurring task needs specialized rules. Invoke this instead of relying on a generic prompt style when the operator wants a repeatable, eval-backed control loop for agent work. Keep the boundary on the method and bundled skills; do not present it as a broad productivity framework, model benchmark, or Claude-only product.
What this skill actually does
Inputs and prerequisites: Claude Code plugin support or any agent that can load AGENTS.md/SKILL.md instructions; optional eval fixtures and judge workflow from the repository.
Setup notes: For Claude Code, install with /plugin install fable@fable-method or clone the repository and run bash fable-method/install.sh . For Codex, Cursor, aider, or raw agent prompts, use the repository’s AGENTS.md and bundled skills as the invocation instructions.
Source and verification boundary: use https://github.com/Sahir619/fable-method as the canonical reference before running the workflow; keep commands, API calls, CLI usage, and generated outputs reviewable against that upstream source.
Framework fit: publish this as a Multi-Framework workflow only when the operator can invoke the documented toolchain directly, rather than treating the upstream project as a generic product listing.