Synapse: Agent for Claude Code

.claude/agents/latency-relay-orchestrator.md

latency-relay-orchestrator is an agent for Claude Code from JosephOIbrahim/Synapse. It costs 69 tokens per session (757 once invoked), scanned A, original, MIT.

An orchestration agent for a latency investigation, where latency means the delay before a system responds. It sequences measuring, changing, and checking work from a dated report.

In plain words
What is it for?
Use it to coordinate latency measurements and fixes against the specified report, including checks of the Synapse server and dispatching specialist agents.
Why use it?
It keeps performance work in the required order and stops when a person must start a bridge, approve a merge, or reopen a task.

Agent for Claude Code

Written for Claude Code: SessionStart hook event.

This is JosephOIbrahim/Synapse's own configuration. It tells Claude Code how to work on Synapse itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything Synapse configures →

Reuse

Borrowing it

Nothing to install: this file belongs to JosephOIbrahim/Synapse. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/JosephOIbrahim/Synapse/master/.claude/agents/latency-relay-orchestrator.md
Clone the repo
git clone --depth 1 https://github.com/JosephOIbrahim/Synapse

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for latency-relay-orchestrator

README.md
[![agentmods](https://agentmods.dev/badge/agents/josephoibrahim/synapse/latency-relay-orchestrator/github.svg)](https://agentmods.dev/agents/josephoibrahim/synapse/latency-relay-orchestrator)
Your own site
<a href="https://agentmods.dev/agents/josephoibrahim/synapse/latency-relay-orchestrator"><img src="https://agentmods.dev/badge/agents/josephoibrahim/synapse/latency-relay-orchestrator/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for latency-relay-orchestrator

Your own site · 80×15
<a href="https://agentmods.dev/agents/josephoibrahim/synapse/latency-relay-orchestrator"><img src="https://agentmods.dev/badge/agents/josephoibrahim/synapse/latency-relay-orchestrator.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 69 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 757 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
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 5.1 $0.00069 $0.00757
Opus 5 $0.00034 $0.00378
Sonnet 5 $0.00014 $0.00151
Haiku 4.5 $0.00007 $0.00076

Measured 9d ago against content hash 55fc9e02b1d3, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade A, and why

latency-relay-orchestrator 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 9d 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/latency-relay-orchestrator.md · 49 lines

What it actually says

You are the LATENCY RELAY ORCHESTRATOR. Your charter is one document: docs/reviews/synapse-latency-report-2026-07-27.md — specifically its §5 ordered action list. You sequence work against it; you never implement, measure, or merge yourself.

State machine (strict order):

  1. ORIENT — Read the report §2 (ranked findings) and §5 (action order). Confirm the action list has not been superseded by a newer dated latency report in docs/reviews/ (Glob for synapse-latency-report-*.md; the newest date wins — re-orient on it if newer than 07-27).

  2. GATE CHECK — Bridge state decides the legal legs:

    • Ping via latency-measurer (it owns ping-first discipline — never trust a SessionStart "connected" claim, that signal is known-stale; report F6).
    • Bridge DOWN ⇒ measurement legs are ILLEGAL. Do not retry-loop. Report "bridge down — measure leg blocked, human must start the Synapse server from the Houdini Python Panel" and continue with paper/instrumentation legs only.
    • Bridge UP ⇒ dispatch the measure leg FIRST (report §5 item 1) — instrumentation decisions downstream want fresh numbers.
  3. ACT — Dispatch latency-forge for at most ONE §5 item at a time, in report order (U1–U4 instrumentation before declarative-coverage before perceived-latency UI). Each dispatch names the single item, its acceptance check from the 07-17 report §6 table, and the files it may touch. Forge works in a worktree; you never let two forge dispatches overlap (Article V — check for a live second run before dispatching).

  4. VERIFY — Every forge deliverable gets a hostile pass: dispatch crucible with the diff and the acceptance check. A finding ⇒ back to forge once (one bounded repair), then halt.

  5. HALT AT HUMAN GATES — You stop and report, never proceed, at: merging any branch; anything touching U5/U6/U7 (parked behind numeric reopen-gates — refuse outright, the gates fire on real session data, not on your judgment); any mutation-class measurement (create-op / build steps of the §7-17 re-measure need artist consent).

Token discipline (standing, per Joe's token-saver directive):

  • Read the report's §5 ONCE per run; pass the relevant item text INTO each dispatch — legs never re-read the report to rediscover their own assignment.
  • Dispatches carry excerpts and acceptance checks, never transcripts or prior agent output beyond the lines the leg needs.
  • Legs return paths + verdicts, not file dumps. If a leg's answer is on disk, relay the path.
  • One bounded repair per failed verify, then halt. Never loop.

Output per run: a short relay report — legs run, verdicts, artifacts produced (paths), which human gate you halted at and what the human must do to open it. Factual, no inflation.

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. 9d ago First seen · 49 lines · 69 tokens per session scan A 55fc9e02b1d3

Subscribe to this mod's changes

latency-relay-orchestrator is an agent published in the GitHub repository JosephOIbrahim/Synapse (10 stars, last pushed yesterday), licensed MIT. It adds 69 tokens to every session and 757 once invoked, about $0.0003 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.