claudecodex

pave-init

Create a PAVE-based Claude Code workflow plugin (a lead skill plus registered role agents) from a stated goal, with adversarial review gates.

1 on GitHubby jinhuang12v2.12.2updated 9 days ago
/

Plugin pave-init from marketplace jinhuang12-plugins in jinhuang12/pave. Commands read from the repo's manifest — 8 days ago.

This plugin runs code on your machine — its hooks fire automatically on events like tool use. The exact commands are listed below; read them before installing.

What it installs

Skills (3)

  • orchestrate

    Turn yourself into a pure orchestrator for any complex task. Group the user's asks into independent streams, investigate them in parallel, and give each stream to a gpt-6.1-sol `orchestrate-worker` that designs and builds it, with a shared gpt-6-astra `adversarial-investigator` per 2-3 workers that

  • pave-evolve

    Revise a delivered PAVE workflow — pave-init included — from recorded defect and usage evidence: a workflow-updater drafts a successor patch, an update-reviewer judges it, the user approves where the user-only changes moved, and the lead applies it in the append-only revision log. Manual-only: use o

  • pave-init

    Create a PAVE-based Codex workflow plugin from a goal the user states. Manual-only: use only after explicit $pave-init:pave-init invocation or selection through /skills. This long campaign clarifies the goal, explores the system, designs and reviews a PAVE graph, and builds and tests the plugin. The

Subagents (10)

  • adversarial-investigator

    Use this agent when you need a rigorous, truth-seeking review of claims, arguments, code logic, architectural decisions, or any content where truth and evidence must be verified rather than taken at face value. This agent combines adversarial framing (actively seeking flaws, contradictions, and unsu

  • forward-tester

    Clean-room test of a generated plugin in a pave-init run — uses its workflow as a normal user would, at real session defaults. Dispatched by the pave-init lead only — do not trigger from an implicit match.

  • node-planner

    Plans one PAVE node for a pave-init run — judges whether one agent can achieve the node's goal, then returns the atomic contract or the child nodes that implement it. Dispatched by the pave-init lead only — do not trigger from an implicit match.

  • orchestrate-worker

    Designs and implements one stream or subtask issued by the orchestrate skill, loads the relevant skills, and loops with its assigned adversarial-investigator until the review PASSes. Dispatched by the orchestrate skill only, as a teammate with model opus passed on the Agent call; do not trigger from

  • pave-material-reviewer

    Material-only adversarial reviewer for pave-init review gates (node plan review, whole-plan gate, final-skill gate). Combines truth-seeking investigation with the pave-init materiality contract. Dispatched by the pave-init lead only — do not trigger from an implicit match.

  • research-delegate

    Fast, focused research sub-agent that receives a single specific question from an investigator and returns evidence-backed findings. Dispatched by the pave-material-reviewer or adversarial-investigator — do not trigger from an implicit match.

  • skill-builder

    Implements one assigned, non-overlapping build unit from an approved PAVE plan in a pave-init run. Dispatched by the pave-init lead only — do not trigger from an implicit match.

  • system-explorer

    Bounded primary-evidence investigation of one system angle for a pave-init run. Writes only its own angle report; never edits the target system. Dispatched by the pave-init lead only — do not trigger from an implicit match.

  • update-reviewer

    Material-only adversarial review of a proposed revision to a PAVE-generated workflow — checks the successor patch against the defect it claims to fix, the live graph at the newest revision, the PAVE spec, and the anti-growth discipline. PASS or REVISE with file:line evidence. Dispatched by a PAVE le

  • workflow-updater

    Drafts a successor revision of a PAVE-generated workflow — pave-init included — from recorded defect and usage evidence. Refactors before it adds, keeps every change generalizable to the workflow's goal, and hands a reviewable patch with its semantic diff back to the lead. Dispatched by a PAVE lead

Hooks (7)

  • SessionStart

    "${CLAUDE_PLUGIN_ROOT}/skills/pave-init/hooks/goal_restate.sh"

  • SessionStart

    "${CLAUDE_PLUGIN_ROOT}/skills/orchestrate/hooks/goal_restate.sh" claude

  • SubagentStart

    "${CLAUDE_PLUGIN_ROOT}/skills/pave-init/hooks/goal_restate.sh"

  • PostToolUse

    "${CLAUDE_PLUGIN_ROOT}/skills/pave-init/hooks/planning-layout-warn.sh"

  • PostToolUse

    "${CLAUDE_PLUGIN_ROOT}/skills/pave-init/hooks/write_for_reader.sh"

  • PostToolUse

    "${CLAUDE_PLUGIN_ROOT}/skills/orchestrate/hooks/speedup_nudge.sh" claude

  • PostToolUse

    "${CLAUDE_PLUGIN_ROOT}/skills/orchestrate/hooks/maintainer_nudge.sh"

View source on GitHub

More from jinhuang12/pave

All 3 plugins →

pave-init-generated workflow: bring the vLLM-Neuron platform plugin fork to parity with upstream GPU vLLM through user-gated, evidence-backed campaigns. Ships the fifteen Neuron skills its campaign stages load inside the plugin (see its README).

16 skills · 6 agents · hooks

1

ammo

claudecodex

Hand-ported PAVE workflow: run a staged, multi-agent campaign that profiles and optimizes Pumice inference latency on your own NVIDIA GPU host, inside a local Pumice checkout, on Claude Code or Codex (see its README).

2 skills · 8 agents · hooks

1

Maintain pave-init?

This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.

claude & codex plugins3 plugins