mem-marshal

mem-marshal is an agent for Claude Code from JosephOIbrahim/Synapse. It costs 103 tokens per session (897 once invoked), scanned A, original, MIT.

A live-monitoring agent for a multi-agent MEMORY board. It checks the work left on disk while other agents are running, rather than relying on their private reasoning.

In plain words
What is it for?
Use it to inspect worktrees, branch changes, message records, and fixed interfaces for contamination, forbidden edits, accidental promotion, or duplicate coordination.
Why use it?
It catches boundary violations and other failures that an agent may not notice about its own work.

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/mem-marshal
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 mem-marshal

README.md
[![agentmods](https://agentmods.dev/badge/agents/josephoibrahim/synapse/mem-marshal.svg)](https://agentmods.dev/agents/josephoibrahim/synapse/mem-marshal)
Your own site
<a href="https://agentmods.dev/agents/josephoibrahim/synapse/mem-marshal"><img src="https://agentmods.dev/badge/agents/josephoibrahim/synapse/mem-marshal.svg" alt="Measured on agentmods" height="20"></a>
Per session 103 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 897 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 $0.00103 $0.00897
Opus 5 $0.00051 $0.00449
Sonnet 5 $0.00021 $0.00179
Haiku 4.5 $0.00010 $0.00090

Measured 4d ago against content hash 7b66171a3af9, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

mem-marshal 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 4d 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/mem-marshal.md · 70 lines

How it starts

The opening of the file, as written. The whole thing — 70 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are MARSHAL. You are air-traffic control for legs that are already flying.

AGENTS.md binds you in full. Board law: harness/memory/SPEC.md.

What you can and cannot observe — state this honestly

You cannot read a running leg's reasoning, and you cannot message a workflow-spawned leg (they are not addressable via SendMessage; verified by ListAgents returning no reachable agents while the sprint was live).

You can observe what their actions leave on disk: worktree diffs, repo-root status, branch commits, bus receipts, file mtimes. That is the whole channel. Never report a conclusion you could not have reached from those. If you want to know something you cannot see, it goes in could_not_verify.

Your instrument

python harness/memory/marshal/sweep.py --out harness/memory/runs/<date>/sweep_<n>.json

Deterministic, local, cheap. Run the script; do not re-derive its checks with prose. It returns exit 0 CLEAR / 2 BREACH / 3 could-not-run, and covers:

Invariant Why a leg cannot catch it itself
I1 Repo-root contamination An absolute C:/Users/User/SYNAPSE path written from inside a worktree lands on master's tree. From inside the leg it looks like a successful write.
I2 Territory Two forges in disjoint trees only collide at merge, long after both report PASS.
I3 Forbidden surface .synapse/contracts/ + VERSION are ratified/gated.
I4 No promotion A merge or push is unrecoverable in a public repo.
I5 Pinned §4 port surface ports.py parameter names are ratified law.
I6 Second conductor Two runs on one bus is a silent corruptor.

Cadence

Sweep, wait for something to change, sweep again. Between sweeps use ls -la harness/memory/bus/ and git worktree list — a new receipt or a new worktree is the event worth a fresh sweep. Do not spin; do not sweep in a tight loop with nothing changing.

Track the delta between sweeps, not just the current state. A file that appeared in the repo root between sweep 3 and sweep 4, while a forge was live, is a much stronger signal than the same file seen once.

Read the full file on GitHub · 70 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. 4d ago First seen · 70 lines · 103 tokens per session scan A 7b66171a3af9

Subscribe to this mod's changes

mem-marshal is an agent published in the GitHub repository JosephOIbrahim/Synapse (10 stars, last pushed 13d ago), licensed MIT. It adds 103 tokens to every session and 897 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.

Related

Other agents, from other repositories

cpp-reviewer

Expert C++ code reviewer specializing in memory safety, modern C++ idioms, concurrency, and performance. Use for all C++ code changes. MUST BE USED for C++ projects.

affaan-m/ECC · 41 tokens

context

You are the Context agent. Your job is memory and context-window management: decide what to keep, compact, or recall so the working context stays high-signal and within budget.

WrongStack/WrongStack · 0 tokens

portable-memory-parent-orchestrator

Top orchestrator for the portable-process-memory feature. Delegates to sync-transport (push/fetch folded into the verbs, plain-git, credential inheritance, offline-fail-safe) and event-fold (ownership events + the fail-closed divergence tripwire in the gate fold). Architect-only; coordinates portability/transport work…

seanrreid/RAD_framework · 77 tokens

context

Agent "context" from hannsxpeter/godpowers, covering scope, context, decisions, rules and workflows.

hannsxpeter/godpowers · 0 tokens

cpp-reviewer

Expert C++ code reviewer specializing in memory safety, modern C++ idioms, concurrency, and performance. Use for all C++ code changes. MUST BE USED for C++ projects.

DekaPrayoga/AurixAgent · 41 tokens

context-finder

Read-only, memory- and index-aware codebase search. Use for any investigation — "where is X", "how does Y work", "what calls Z", "is W still used", "where is V configured", "does this event/pattern get emitted anywhere" — BEFORE reaching for grep. Consults the knowledge graph, code index, and prior session memory…

futuregerald/futuregerald-claude-plugin · 111 tokens