wave AGENTS.md

A project instruction file for Wave, a Go program that coordinates multiple AI coding agents and command-line tools. It explains the project's architecture, development rules, testing expectations, and security constraints.

In plain words
What is it for?
Use it when developing Wave's agents, pipelines, command-line integrations, contracts, security checks, or tests.
Why use it?
It gives an agent the context needed to change a multi-agent system without breaking its contracts or runtime behavior. It also requires problems found during work to be fixed and validated.

Instructions file for CodexOpenCode

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add instructions/re-cinq/wave/agents-md
Clone the repo
git clone --depth 1 https://github.com/re-cinq/wave

Made for: Codex, OpenCode.

Per session 4,725 This file is loaded in full into every session.
When invoked 4,725 The same file — it is already loaded in full.
Security scan B 1 finding. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.04725 $0.04725
Opus 5 $0.02363 $0.02363
Sonnet 5 $0.00945 $0.00945
Haiku 4.5 $0.00473 $0.00473

Measured yesterday against content hash 796e02e2d5f9, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade B, and why

wave AGENTS.md scanned grade B with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Reads agent configuration directoriesmediumAgent snooping

.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.

6. Check for leaked files: `gh pr diff <N> --name-only | grep -E "^\.claude/|^\.agents/artifacts/|^\.agents/output/"`
AGENTS.md · 303 lines

How it starts

The opening of the file, as written. The whole thing — 303 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Wave Development Guidelines

Wave is a multi-agent pipeline orchestrator written in Go that wraps Claude Code and other LLM CLIs via subprocess execution. It composes personas, pipelines, contracts, and relay/compaction into a continuous development system.

ACCOUNTABILITY — YOU FOUND IT, YOU FIX IT

URGENT — NON-NEGOTIABLE: This is the single most important rule in this file. It survives context compaction. Re-read it if you are unsure.

If you discover a problem — any problem — you own it. Fix it immediately.

There is NO concept of "pre-existing issue" in this project. If you touched it or saw it break — fix it.

If a validation step reveals issues in files you didn't modify — fix those too.

Never argue about whether something is your responsibility.

Critical Constraints

  1. Single static binary — no runtime dependencies except adapter binaries
  2. Test ownership — every failing test is YOUR concern. Fix or delete (with justification), never ignore. Changes to personas, pipelines, contracts, or meta-pipelines require go test ./...
  3. Security first — all inputs validated, paths sanitized, permissions enforced
  4. Constitutional compliance — navigator-first architecture, fresh memory at step boundaries, contract validation at handovers, ephemeral workspace isolation, observable progress events
  5. Observable execution — structured progress events for monitoring
  6. No backward compatibility constraint during prototype phase — move fast, let tests catch regressions
  7. No t.Skip() without a linked issue. Delete tests only with clear justification

How Wave Works at Runtime

Each pipeline is a topologically-sorted DAG of steps. For every step:

  1. Workspace creation — an ephemeral worktree is created under .agents/workspaces/<pipeline>/<step>/. Steps can share workspaces via workspace.ref. Mounts support readonly/readwrite modes
  2. Artifact injection — outputs from prior steps are injected into .agents/artifacts/ before execution begins. The system validates existence, enforces optional/required semantics, and checks schemas if ref.SchemaPath is specified
  3. Runtime CLAUDE.md assembly — a per-step CLAUDE.md is generated from four layers:
    • Base protocol preamble (.agents/personas/base-protocol.md)
    • Persona system prompt (role, responsibilities, constraints)
    • Contract compliance section (auto-generated from step contract schema)
    • Restriction section (denied/allowed tools, network domains from manifest permissions)
  4. Adapter execution — the persona runs in isolated context with fresh memory (no chat history inheritance)
  5. Contract validation — step output is validated against its contract (json_schema, typescript_interface, test_suite, markdown_spec, format) before marking the step successful. Hard failures block; soft failures log warnings

Read the full file on GitHub · 303 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 303 lines · 4,725 tokens per session scan B 796e02e2d5f9

Subscribe to this mod's changes

wave AGENTS.md is an instructions file published in the GitHub repository re-cinq/wave (20 stars, last pushed 4mo ago), licensed MIT. It adds 4,725 tokens to every session, about $0.0236 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other instructions, from other repositories

coordinate-agents AGENTS.md

Instructions for hogancv/coordinate-agents, covering repository instructions for ai contributors, canonical identity, repository map, required checks and change rules.

hogancv/coordinate-agents · 1,451 tokens

claude-multi-agent-architecture CLAUDE.md

Instructions for mnzralee/claude-multi-agent-architecture, covering quick start (every session), common commands, when rules conflict, multi-agent orchestration and reasoning effort.

mnzralee/claude-multi-agent-architecture · 1,685 tokens

antigravity-agents GEMINI.md

Instructions for rafaelghif/antigravity-agents: Read AGENTS.md before acting. It is the single source of truth for always-on policy. Consult .agents/brain/activecontext.md for active session state and .agents/brain/memory.md & rules.md for permanent project architecture, stack profile, and user preferences across all…

rafaelghif/antigravity-agents · 68 tokens

claude-multi-agent-architecture AGENTS.md

Instructions for mnzralee/claude-multi-agent-architecture, covering agents.md - cross-tool agent configuration, project overview, conventions, code style (illustrative defaults; adapt to your stack) and git.

mnzralee/claude-multi-agent-architecture · 1,259 tokens

antigravity-agents AGENTS.md

Instructions for rafaelghif/antigravity-agents: You are an L9 Engineer with System-2 Test-Time Compute (TTC) & strict verification. Simulate paths logically; validate assumptions via static analysis.

rafaelghif/antigravity-agents · 1,769 tokens

cctop AGENTS.md

Instructions for st0012/cctop, covering agents.md - development guide for cctop, scope and sources, required development rules, driver workflow and github and pr rules.

st0012/cctop · 2,107 tokens