Iteratively generate, critique, and revise engineering plans until blocking issues are removed.
npx skills add https://github.com/zeikar/hyperclaude --skill hyper-plan-loop
Please help me install this Agent Skill. Skill: hyper-plan-loop Source: https://github.com/zeikar/hyperclaude/tree/main/skills/hyper-plan-loop Command: npx skills add https://github.com/zeikar/hyperclaude --skill hyper-plan-loop
This Skill removes the uncertainty of whether a generated engineering plan is correct and implementable by repeatedly running Codex critique and revising the plan until blocking issues are gone.
Use /hyperclaude:hyper-plan-loop add OAuth login to the API to generate a plan, review it with Codex, and revise it until it’s clean.
High-intent search queries and answers about installing and using this skill.
Autonomous plan hardening works by spawning a live planner teammate that generates a plan, then repeatedly running Codex plan-review and revising the artifact until no blocking findings remain.
Iterative plan critique uses a live planner teammate spawned once and reused across revise rounds, retaining its planning context to ensure continuity throughout the plan hardening workflow.
Meaning-based gating treats Codex findings as blocking only if they affect plan-level correctness, ignoring generic severity labels to focus on issues that actually prevent implementation.
Yes, this resume-aware plan hardening workflow requires an experimental agent-teams setup to spawn the live planner teammate and execute the codex-bridge plan-review iterations.
Plan artifacts are written to a canonical plan path within the .hyperclaude/plans directory, enabling artifact-driven validation, review, and workflow continuity across multiple revise rounds.
Yes, optional research artifact inlining is supported through slug matching, allowing you to integrate external research directly into the plan during the iterative critique and revision process.
zeikarExecutes a plan end-to-end by looping implement, code-review, and targeted fixes until blocking findings resolve.
CityBear3Execute approved implementation plans in isolated workspaces with verification gates.
AGMO-IncValidate plans through iterative critic reviews and acceptance criteria.
choms0521Coordinate Codex and Claude planners and critics to refine plans.
