miaodx/intuitive-flow/agent-planning-loop
Claude Code and agent skill by miaodx/intuitive-flow
- community
Checking that the SKILL.md file is still in the repository.
Review unsettled plans with independent scouts when the user requests multi-agent planning or alignment.
Install
npx vskill@latest install miaodx/intuitive-flow/agent-planning-loopSKILL.md
Agent Planning Loop
Run a bounded planning debate when requested and return one judged recommendation. Implementation and approval happen after the loop, and questions for the user go through the main session, not the scouts. Preserve the full intended scope of a supplied plan unless the user selects a subset or a material decision blocks honest full-plan execution.
Plan artifact ownership
For a normal planning request, the canonical artifact is the target repository's existing plan path. When the repository convention is docs/plans/, use one docs/plans/<slug>.md file and reconcile the loop's accepted decisions into that file. A planning loop reviews or updates the canonical plan; it does not create .planning/, ROADMAP.md, STATE.md, phase CONTEXT.md, or a GSD PLAN.md as a side effect.
Mentioning GSD, agents, or a planning loop does not itself authorize a GSD handoff. Stop at the canonical plan checkpoint for a planning-only request. A GSD artifact may be generated only when the user requests implementation/GSD handoff or an existing GSD phase already owns execution, and the named GSD workflow has actually been invoked.
Set a brief charter: goal, non-goals, context, permitted worker actions, and stop condition. Use scouts when preparing worker prompts and judging their evidence. Main-session judgment owns the recommendation.
Default to at most two rounds: discovery/challenge, then a narrowed follow-up only if it can change the recommendation. Stop sooner when a plan has clear acceptance and proof, the remaining items are implementation defaults, or a user decision blocks progress. A deeper autonomous loop requires the user's request.
Output
Return a review packet, not a transcript.
Include: planning loop status, what ran, rejected/parked items, one recommended plan with scope/non-goals/acceptance/verification/risk, alternatives only when material, user decisions, Plan artifact:, Recommended next action:, and Shortcut:.
For plan-file loops, the Recommended next action should normally execute or preflight the whole plan through the appropriate route. It may name the first phase as the starting point, but should not make that phase the whole approved scope unless the user asked for that narrowing or a stop gate blocks the rest.
For a repository plan-file loop, Plan artifact: must identify the canonical plan path (normally docs/plans/<slug>.md), not a review log or a newly-created GSD phase file.
If no material plan remains, say so directly and explain what evidence caused the stop; a short packet beats weak alternatives. Still include Recommended next action and Shortcut; use park/none when there is genuinely no useful next step.
Approval Handling
If the user approves the review packet, the loop is done unless their approval changes scope. Route to the single Recommended next action. Treat short replies such as LGTM, approve, sounds good, or do it as approval for that action and preserve the packet's plan artifact, scope, verification, user decisions, parked items, and stop condition.
If the user pushes back on part of the recommendation, treat that as new charter input. Rerun only the affected scout stage, not the whole loop.
Security
- Security scanPassed, 100 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/agent-planning-loop/SKILL.md
Popularity
- Repository stars
- 48
- Repository forks
- 3
- 7-day trend score
- 19
Stars and forks are for the whole GitHub repository, not this skill alone. 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/miaodx/intuitive-flow/agent-planning-loop)