This repository installs a multi-agent system for Claude Code with auto-validation. The main conversation IS the orchestrator. When the user gives you a task, follow the orchestration flow below —…
When you run inside a nodeterm canvas session, this node may be linked to other agent nodes (Claude, Codex or Gemini) or sticky notes by a context-link edge. You can…
Create a pull request for the current branch following project conventions. Validates branch name, runs pre-push checks, gates before push, and produces a structured PR body.
Phase 3 of the Medium-path workflow. Use after plan produced PLAN.md. Iterates slices, spawning coder+verifier in chain mode (with retry-on-fail) or multiple coders in parallel for disjoint slices. Handles HITL gates for high-risk paths.
Phase 1 of the Medium-path workflow. Use when triage classifies a task as medium or substantial. Delegates to the scout sub-agent to gather codebase context + Neurox prior decisions, then returns a structured exploration envelope to be passed to the plan skill.
Relentless one-question-at-a-time discovery for non-trivial features. Use BEFORE writing a PRD, proposal, or plan. Stops vibe coding. Forces shared design concept with the user.
First-time project onboarding. Scans codebase, identifies stack/conventions/architecture, and saves knowledge to Neurox for all future sessions. Run once per project.
Phase 2 of Medium and Substantial-path workflows. Invokes tech-planner to produce PLAN.md. On SUBSTANTIAL path, this is the SINGLE HITL gate by default — blocks for explicit user approval before /skill:build (showing proposal.md + SPEC.md + PLAN.md together). On medium path, only blocks if new_dependencies non-empty or any slice has requires_hitl=true (current Sprint 2 behavior).
Formal Product Requirements Document for stakeholders. Use when a feature needs business sign-off before engineering. Produces user stories, acceptance criteria, NFRs, and out-of-scope sections.
Substantial-path proposal skill. Invokes product-planner to write a 1-page proposal.md. By default AUTO-CONTINUES to /skill:specify (no gate). Only blocks for user approval when SKYNEX_HITL=strict.
Deep web research on external topics — libraries, APIs, prior art, documentation. Use when local context is insufficient and internet lookup is needed.
Research mode synthesis protocol. Called by the main model after 3 parallel research agents (neurox, web, code) return their envelopes. Synthesizes a final verdict with source attribution.
Task-creation mode flow. Sequential 4 steps — grill, decompose, HITL gate, Jira creation. Activated by /skynex:task. Main model is the executor; no sub-agents.
Substantial-path specification skill. Invokes product-planner + architect IN PARALLEL, merges into unified SPEC.md. By default AUTO-CONTINUES to /skill:plan (no gate). Only blocks for user approval when SKYNEX_HITL=strict.
Plain text files in a repository that tell a coding agent how the project works: commands to run, conventions to follow and things to avoid. CLAUDE.md, AGENTS.md, cursor rules and skills are the common kinds.
CLAUDE.md or AGENTS.md?
CLAUDE.md is read by Claude Code. AGENTS.md is an open format that Codex, Cursor and other agents read. Many projects keep one and point the other at it.
What is a skill?
A folder with a SKILL.md that describes one capability, such as filling PDFs or reviewing code. The agent loads it only when the task calls for it.
Can I search my own team's files too?
Your agents already can, over MCP, limited to the files you're allowed to read. Searching them from this page is coming.