richkuo/rk-skills/fable-validate-fableplan-loop
Claude Code and agent skill by richkuo/rk-skills
- community
Checking that the SKILL.md file is still in the repository.
Use when the user asks to validate a GitHub issue with Fable 5.1, always plan it with Fable 5.1, and autonomously drive it to a reviewed PR in one shot — "fable-validate-fableplan-loop", "fable validate, fable plan, and work on #N", "fully automate #N with fable validation and an unconditional fable plan". Runs fable-validate, auto-applies its update-issue edits when the verdict calls for it, has fableplan produce and post a Fable 5.1 implementation plan for EVERY issue (no score gate — unlike…
Install
npx vskill@latest install richkuo/rk-skills/fable-validate-fableplan-loopSKILL.md
fable-validate-fableplan-loop
Chain fable-validate → (conditional) update issue → fableplan → work-on-issue-loop into one autonomous run.
This is fable-validate-loop with the score gate removed: fableplan always runs, whatever the validated complexity score, and there is no "skip when the score is below 75" rule. When skipping the plan for lower-band issues is fine, use fable-validate-loop.
Do not skip or reorder the chain. Validation gates planning, and the plan gates implementation. This variant has no sanctioned skip: every step runs, and only the "wait for the user's reply" moments are replaced by the decision rules in the cited steps.
Input
Same as fable-validate-loop, including the optional targetBranch.
Steps
Follow fable-validate-loop steps 1 through 6 with these changes:
Step 1 (fable-validate) applies unchanged.
Step 2 (scope gate) applies unchanged, with fable-validate-loop step 2's STOP table.
Step 3 (update-issue edits): apply them per fable-validate step 5 / validate-issue step 11; the stacked Validated with LLM: … attribution line uses the harness suffix fable-validate-fableplan-loop.
Step 4 (fableplan): there is no score gate — fable-validate-loop's score gate, safety carve-out, and top-band note do not apply; fableplan runs for every issue that passed the step-2 scope gate, whatever the validated score. Instruct fableplan to use the harness suffix fable-validate-fableplan-loop in the posted comment's attribution footer. The rest of the step applies unchanged: fableplan's planning-phase-only invocation, and the validation verdict handed to the planner (verified/refuted claims, the Optimal-direction note when architecture was Underspecified, 5c concerns), with the scratchpad kept for step 5.
Step 5 (handoff) applies unchanged — including that deviations follow work-on-issue step 2's plan-deviation policy and must each be named in the PR body.
Step 6 (report) applies unchanged. Cap the whole report at 55 words and 5 sentences, plain simple English in ASD-STE100 — apply the Response Style rules in CLAUDE.md/AGENTS.md.
Red Flags — STOP
fable-validate-loop's Red Flags table applies. Read its "only sanctioned skip is the step-4 score gate" wording as "no sanctioned skip at all": a score below 75 never skips fableplan here.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/fable-validate-fableplan-loop/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/richkuo/rk-skills/fable-validate-fableplan-loop)