Synapse AGENTS.md

A set of rules for building SYNAPSE, requiring agents to support claims with repeatable evidence. It also defines how missing or unavailable capabilities must be reported.

In plain words
What is it for?
Use it when developing SYNAPSE or coordinating agents that need traceable evidence for code changes, tests, and results.
Why use it?
It reduces the risk of agents claiming that work succeeded when it was not actually done.

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/josephoibrahim/synapse/agents-md
Clone the repo
git clone --depth 1 https://github.com/JosephOIbrahim/Synapse

Made for: Codex, OpenCode.

Per session 2,660 This file is loaded in full into every session.
When invoked 2,660 The same file — it is already loaded in full.
Security scan A 0 findings. 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 $0.02660 $0.02660
Opus 5 $0.01330 $0.01330
Sonnet 5 $0.00532 $0.00532
Haiku 4.5 $0.00266 $0.00266

Measured 2d ago against content hash b5a977f9477e, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

Synapse AGENTS.md scanned grade A with 0 findings 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 2d ago.

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.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

AGENTS.md · 272 lines

How it starts

The opening of the file, as written. The whole thing — 272 lines — stays where its author put it; the contents beside it link to each section on GitHub.

AGENTS.md — How agents work on SYNAPSE

Law for every subagent, workflow leg, and orchestrator that touches this repo. CLAUDE.md says what SYNAPSE is. This says how you are allowed to build it. If the two disagree, CLAUDE.md wins and you file the contradiction.


0 · Prime directive

Produce evidence, not confidence.

Every claim you make must be traceable to something a hostile reader can re-run: a file path with a line number, a probe artifact, a test name, a command and its output. A sentence with no producer path is a rumor.

The single most expensive failure mode in this repo is not a bug. It is a green receipt for work that did not happen. Everything below exists to make that failure structurally hard.


1 · The Seven Laws

Law 1 — Honest seam

A capability whose substrate is not installed reports UNAVAILABLE with the reason. Never a fabricated SUCCESS. Never a guessed BLOCK. Never a placeholder that returns the shape of an answer without the answer.

A stub that returns SUCCESS is worse than no stub. No stub fails loud; a lying stub fails silent and takes a test suite hostage with it.

Law 2 — Absence has a shape

Absence is a measured fact, not a missing feature — and how you degrade depends on which side of the seam the absent substrate sits. See §2.

Law 3 — Runtime is truth

A live-introspected surface beats a doc claim. A probe artifact beats a pinned constant. If the doc and the runtime disagree, the runtime is right and the doc is a bug you now own.

Before emitting any hou.* / pdg.* / pxr.* symbol you are not certain of, call synapse_scout. Phantom APIs are this repo's #1 historical failure class.

Law 4 — Unmeasured renders UNKNOWN

No estimates. No "should be roughly." An unmeasurable goalpost is written UNKNOWN and the leg does not go green on it.

Law 5 — Receipts or it did not happen

Every leg posts a receipt to its board's bus/ before it reports done. The receipt carries: what you touched, the command you ran, the artifact path, the verdict, and what you could not verify. The last field is not optional, and "none" is rarely the honest answer.

Read the full file on GitHub · 272 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. 2d ago First seen · 272 lines · 2,660 tokens per session scan A b5a977f9477e

Subscribe to this mod's changes

Synapse AGENTS.md is an instructions file published in the GitHub repository JosephOIbrahim/Synapse (10 stars, last pushed 11d ago), licensed MIT. It adds 2,660 tokens to every session, about $0.0133 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.