Instructions file CodexOpenCode
Project-specific instructions for DevRails-26, a framework for agent-driven software development. It deploys skills, shared memory, workflows, protocols, and validators into other projects.
Agent
A review agent that checks whether a project's Memory Bank follows its documentation rules and contains the required files, links, and metadata.
Agent
You are a delegated subagent. Your job is to scan a bounded part of the repo and produce a structured report.
Agent
Review the assigned task's actual implementation change surface against the provided target, governing evidence, and scope. Read-only.
Agent
A review agent for implementation plans and structured task records created by another agent. It checks task organization, dependencies, required fields, outputs, tests, and verification steps.
Agent
A review agent that checks whether another agent's Memory Bank fully covers the project's requirements and planned work.
Agent
A review agent that checks project notes and code for security risks, such as weak access controls, injection flaws, exposed secrets, and unsafe logging.
Command
Add the cheapest sufficient regression coverage inside one existing in-progress indexed task.
Command
Bounded C4 and architecture review of one feature's task-planning surface for a fresh Reviewer.
Command
Run the complete unattended DevRails workflow from authoritative product input to a terminal sequential JSON task queue.
Command
Sequentially execute an already reviewed and strict-ready indexed product JSON task queue to a terminal state.
Command
Facilitated ideation report before Product Brief; AI coaches, operator intent drives.
Command
Create or update a concise Product Brief and initial project glossary.
Command
Adaptively clarify and maintain the Project Constitution governing layer.
Command
Build a compact Memory Bank/codebase Context Read Manifest for another agent from a delegated Explorer session; use for broad discovery when direct reads are not already cheaper.
Command
Use first for creative work—technical or everyday—when a brief is incomplete or intent is implicit and success depends on taste, voice, feeling, or human experience. Also trigger when literal compliance may lose the point. Do not wait for explicit creative wording. Skip factual lookup, mechanical or exact tasks, and…
Command
Diagnose one task failure read-only and route evidence-backed recurrence prevention.
Command
Adaptively resolve unknowns and contradictions before planning or coding.
Command
Execute one indexed JSON task within its semantic and hard runtime boundaries, then produce tier-routed evidence and handoff.
Command
Bounded semantic triage and repair routing for one feature.
Command
Plan or reconcile one product feature into canonical SDD coverage, an implementation plan, and complete JSON task cards.
Command
A command that loads a project's Memory Bank, a set of documents containing its product details, requirements, design, and working rules, into the agent's context.
Command
A project-first search command for finding relevant coding-agent skills, first among skills already installed in the project and then in a marketplace.
Command
Convert an accepted Foundation Dev Path decision into the minimum FT-000 JSON queue, or prove an existing brownfield baseline.