npx agentspec-cli install 4a9c6009-17cf-4800-abfd-a1f9307edcf6 --file-type agents-md→ .agents/
# AGENTS.md ## Project Overview OxiBelt is a Rust-based reverse proxy and WAF project. This repository is organized as a monorepo. The main reverse proxy implementation lives under `source/`. Tests live under `tests/`. Technical specifications and configuration documentation live under `docs/`. Deployable Helm and observability assets live under `deploy/`. TypeScript-based DevOps and CI support code should live under `devops/` when present. The project should be testable locally and in CI using Docker-based environments. ## Repository Structure - `Cargo.toml` - Rust workspace, shared package metadata, and dependency policy. - `source/` - Integrated data-plane, Admin, WAF, and Person Proof runtime crate. - `source/apps/` - Independently packaged Gateway Controller, CLI, keysigner, and netport binaries. - `source/crates/` - Shared external-control protocol, HTTP, and deployment-diagnostics crates. - `source/assets/` - Canonical build-validated assets embedded in the runtime. - `source/src/` - Core application source code. - `source/src/proxy/` - HTTP, HTTP/3, stream, WebSocket, and WebTransport proxy behavior. - `source/src/waf/` - OxiRule, CRS compatibility, body scanning, Person proof, and WAF behavior. - `source/src/config/` - Configuration validation modules. - `source/config/oxibelt.toml` - Example or default OxiBelt configuration. - `source/ops/Dockerfile.alpine` - Alpine-based Docker image for OxiBelt. - `tests/rust/` - Rust integration tests and repository-level checks. - `tests/docker/` - Docker test services, protocol probes, database fixtures, and performance probes. - `tests/scripts/` - Test, build, integration, performance, and WebDriver orchestration scripts. - `docs/` - Technical specifications, configuration guides, rule documentation, and behavior references. - `ui/person-proof/` - Person proof challenge UI assets and build scripts. - `kernel-extension/` - Linux edge deployment tuning templates and verification helpers. - `deploy/` - Deployable Helm charts and observability bundle assets. - `devops/` - TypeScript-based DevOps and GitHub Actions support code when present. - `.github/workflows/` - GitHub Actions workflows. ## Contributor Guidance `CONTRIBUTING.md` is the source of truth for contributor workflow, security requirements, pull request checks, and commit-message format. Use these sections before making or reviewing changes: - [Contribution Workflow](CONTRIBUTING.md#contribution-workflow) - [Commit Messages](CONTRIBUTING.md#commit-messages) - [Security Requirements](CONTRIBUTING.md#security-requirements) - [Pull Request Checklist](CONTRIBUTING.md#pull-request-checklist) If this file and `CONTRIBUTING.md` diverge on workflow, security, testing, documentation, or Conventional Commits requirements, follow `CONTRIBUTING.md` and update this pointer file only when agent-specific orientation changes. ## Agent Commit Message Guidance Commit messages must contain portable, repository-relevant context. Do not include session-specific command aliases, absolute host paths, or local-only environment data and artifacts. For example, do not record the availability or use of `docker-rootful` or cite files under `.agents/temp`; describe the portable result instead, such as whether performance benchmarks were run. When additional context is useful, prefer stable sources that readers can access publicly, such as tracked repository files, public GitHub projects, issues, pull requests, commits, published security advisories, and official documentation. If no suitable public source exists, explain the necessary context inline without citing inaccessible local material. Sanitize the description and do not expose secrets, personal data, or undisclosed vulnerability details. ## Codex Security Advisory Guidance Follow the [Security Policy](SECURITY.md) for every validated, high-confidence, report-worthy finding from Codex Security Cloud or a local Codex Security scan. Use one GitHub repository Security Advisory per finding and keep the disclosure state fail-closed: - If a supported fix or actionable mitigation is not yet available, submit only a private vulnerability report to the repository advisory inbox and keep it in `triage` or `draft`. Do not substitute a public issue, discussion, pull request, or revealing commit message. - Once the supported fix or actionable mitigation is verified and disclosure is permitted by the Security Policy, publish the repository Security Advisory. - Preview the exact advisory payload and obtain explicit approval for each submission or publication. Before writing, verify the canonical repository, immutable source revision and finding locations, duplicate status, authenticated identity, permissions, and intended visibility. Read the advisory back before reporting success. - Do not request a CVE identification number unless the user explicitly asks for one. - If the available credentials or tooling cannot preserve the required private or published state, stop and report the blocker without changing disclosure channels.
Rust on Nails Agents Guide
This repository mixes several deliverables that move at different speeds. To keep things tidy, we split the work into three "agents" that each own one slice of the Rust on Nails experience. Use this guide to decide where changes belong and how to validate them before opening a PR.
by AgentHub Crawler
Autonomous Agent Operations Guide for eo-processor
This document defines how autonomous engineering agents should operate within the `eo-processor` repository. It covers: 1. Core mission and constraints 2. Repository awareness 4. How to add functionality (Rust + Python)
by AgentHub Crawler
Using Vite+, the Unified Toolchain for the Web
This project is using Vite+, a unified toolchain built on top of Vite, Rolldown, Vitest, tsdown, Oxlint, Oxfmt, and Vite Task. Vite+ wraps runtime management, package management, and frontend tooling in a single global CLI called `vp`. Vite+ is distinct from Vite, but it invokes Vite through `vp dev` and `vp build`.
by AgentHub Crawler
Sign in to leave a comment.
AGENT.md
This repository is an implementation of OpenAI Symphony as described in [docs/symphony-spec.md](docs/symphony-spec.md). The upstream specification is the source reference for architecture, layering, orchestration behavior, runtime behavior, workflow loading, workspace lifecycle, tracker integration boundaries, and observability expectations.
by AgentHub Crawler