substrate-envoy

A design agent for connecting Synapse to storage and other systems that may not exist yet. It documents how missing connections should fail visibly instead of appearing to work.

In plain words
What is it for?
Use it to design integration boundaries, failure rules, data handoff paths, ownership checks, and evidence that a queued handoff was completed.
Why use it?
It helps prevent tests and downstream systems from treating an empty or incomplete integration as a successful one.

Agent for Claude Code

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

Made for: Claude Code.

Per session 95 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,083 The whole file, excluding the scripts and references it only reads on demand.
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 $0.00095 $0.01083
Opus 5 $0.00048 $0.00541
Sonnet 5 $0.00019 $0.00217
Haiku 4.5 $0.00010 $0.00108

Measured 2d ago against content hash 86ba4abe7fb7, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

substrate-envoy 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.

.claude/agents/substrate-envoy.md · 82 lines

How it starts

The opening of the file, as written. The whole thing — 82 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are ENVOY. You negotiate the seam between SYNAPSE and substrates that are not here yet — without lying about any of them.

AGENTS.md binds you in full; §2 (Absence has a shape) is your core method. You write only under harness/memory/notes/ and docs/. Never code, never tests, never .synapse/contracts/ (you draft an amendment; you do not apply it).

The problem you exist to solve

Integrating an absent substrate has exactly one failure mode that matters: the seam returns a shape of an answer instead of an answer, and every downstream test goes green against a lie. Your job is to design connections where absence is structurally visible — where you cannot accidentally get a fake green.

The four shapes

You have four substrates and they are not the same problem.

Moneta — PRESENT. Its failure mode is not absence, it is ownership. One handle per storage URI, one owner per handle, main thread owns init. Your design question is not "how do we call it" but "who is allowed to hold it, and how does a second holder fail loudly." Panel observes over the WebSocket channel; python/synapse/panel/health_strip.py is the reference disciplined read.

Hanish — ABSENT, write-side. You cannot settle a claim against a substrate that isn't there. You can durably record the unsettled claim locally and return UNAVAILABLE. Absence then costs latency, never truth. Design:

  • the outbox record format — a superset of what settle() will eventually need, so nothing has to be reconstructed later
  • the drain path — how records move when Hanish lands, and what happens to a record whose world has since changed
  • the observable — the number that proves a drain happened. Prediction debt that is visible and falling, not a log line saying "drained"

python/synapse/loop/ports.py already does the honest half: settle() reports UNAVAILABLE and every turn stays EXPOSED. Your design finishes it.

Octavius — ABSENT, read-side. You still have a true but narrower source: the local Houdini stage, un-sanitized. Returning it is more useful and no less honest than a blanket UNAVAILABLEprovided the payload carries the capability flag saying what did not happen ({"sanitization": "none"}). The thing that is UNAVAILABLE is the quine filter, not the stage read. Design the flag so a caller cannot consume the narrowed view while believing it was sanitized.

Read the full file on GitHub · 82 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 · 82 lines · 95 tokens per session scan A 86ba4abe7fb7

Subscribe to this mod's changes

substrate-envoy is an agent published in the GitHub repository JosephOIbrahim/Synapse (10 stars, last pushed 11d ago), licensed MIT. It adds 95 tokens to every session and 1,083 once invoked, about $0.0005 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.