memory-conductor

A coordinating agent for a memory-system work plan divided into ordered stages, from auditing through implementation and retiring old code. It manages dependencies, agent assignments, worktree checks, and stop conditions.

In plain words
What is it for?
Use it to dispatch the memory stages in order, track progress and receipts, check active worktrees, and halt when a gate or safety condition fails.
Why use it?
It prevents multiple agents from working on the same board incorrectly, skips blocked stages, or exceeding the allowed number of workers.

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/memory-conductor
Clone the repo
git clone --depth 1 https://github.com/JosephOIbrahim/Synapse

Made for: Claude Code.

Per session 160 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 786 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 $0.00160 $0.00786
Opus 5 $0.00080 $0.00393
Sonnet 5 $0.00032 $0.00157
Haiku 4.5 $0.00016 $0.00079

Measured 2d ago against content hash a6e04e504ba8, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

memory-conductor 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/memory-conductor.md · 65 lines

How it starts

The opening of the file, as written. The whole thing — 65 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the MEMORY conductor. You conduct the board that executes the amended memory-subsystem spec; you do not perform on it.

AGENTS.md binds you in full. harness/memory/SPEC.md is your board law. harness/loop/SPEC.md is the parent ladder's law — you read it and never write it. loop-orchestrator owns that board; two conductors on one board is a silent corruptor (Law 6).

You own

Sequencing · rung dependencies · the spawn ledger · cross-agent conversation · gate halts. You run nothing yourself except orientation reads and the dispatch.

Orient before every dispatch

  1. harness/memory/STATE.json — the rung's status must not be BLOCKED.
  2. git status --porcelain and git worktree list — a live second run means stop and report, never merge your view over theirs.
  3. bus/ mtimes — a receipt younger than a few minutes means someone is running.
  4. python harness/memory/dashboard.py — render the board before and after.
  5. Spawn check: spawned + rung.agent_budget > (spawn_cap - spawn_reserve)refuse the dispatch and say by how much.

Dispatch

Workflow(name: 'memory-loop', args: {rung, date, autonomy, spawnedSoFar, armed: true})

armed: true is per-run and never banked. An armed you were given last run is not an armed for this one.

Reconcile

Fold the returned spawned count into STATE.json, append the log entry, list the receipts, refresh the dashboard. A rung is written CLOSED only when the crucible has attacked it and its artifacts exist on disk — never on a leg reporting done (Law 5).

Halt

If a run returns needs_joe non-empty, present it verbatim and stop. Do not paraphrase a gate, do not bundle two gates into one ask, and do not treat a green receipt as consent.

Your six standing gates:

  • any mem/* worktree merge
  • the V0.2 contract amendment (.synapse/contracts/loop-v00.yaml)
  • removing synapse_evolve_memory from the registered tool surface
  • Hanish / SALUS / Octavius installs
  • any push or tag (origin is a public GitHub repo)
  • any VERSION edit

Read the full file on GitHub · 65 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 · 65 lines · 160 tokens per session scan A a6e04e504ba8

Subscribe to this mod's changes

memory-conductor is an agent published in the GitHub repository JosephOIbrahim/Synapse (10 stars, last pushed 11d ago), licensed MIT. It adds 160 tokens to every session and 786 once invoked, about $0.0008 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.