context-manager

context-manager is an agent for coding agents from niels-emmer/myace. It costs 32 tokens per session (543 once invoked), scanned A, original, MIT.

A context-planning specialist for long-running or multi-agent tasks. An agent's context is the working information it can see during a task.

In plain words
What is it for?
Choosing which decisions, file excerpts, and compact error summaries to pass between task stages or agents, while tracking longer-term state externally when available.
Why use it?
It prevents useful decisions from being buried under old files, repeated errors, and unnecessary conversation history.

Agent

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/niels-emmer/myace/context-manager
Clone the repo
git clone --depth 1 https://github.com/niels-emmer/myace

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 context-manager

README.md
[![agentmods](https://agentmods.dev/badge/agents/niels-emmer/myace/context-manager.svg)](https://agentmods.dev/agents/niels-emmer/myace/context-manager)
Your own site
<a href="https://agentmods.dev/agents/niels-emmer/myace/context-manager"><img src="https://agentmods.dev/badge/agents/niels-emmer/myace/context-manager.svg" alt="Measured on agentmods" height="20"></a>
Per session 32 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 543 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 $0.00032 $0.00543
Opus 5 $0.00016 $0.00271
Sonnet 5 $0.00006 $0.00109
Haiku 4.5 $0.00003 $0.00054

Measured yesterday against content hash 7bef08d448d1, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

context-manager 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 yesterday.

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.

collections/additional/ai-engineering/agents/context-manager.md · 28 lines

How it starts

The opening of the file, as written. The whole thing — 28 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Makes an explicit call about what belongs in context for the next step of a long-running or multi-agent task, instead of letting context accrete by default.

Responsibilities

  • For a task spanning multiple stages or agents (e.g. this collection's orchestratorbuilderverifier pipeline, if composed), decide what each stage actually needs: recent decisions and their rationale, not the full history that produced them; the relevant file excerpt, not the whole file.
  • Compact failures before they're passed forward — a failed attempt's root cause and what was already ruled out, not its full raw error output repeated on every retry. See the agent-design-principles skill's note on compacting errors into context.
  • Keep long-running state in one place external to any single agent's context (the project's memory files, if the memory-system skill is in use) rather than relying on any one agent's window to carry it across a session boundary — a resumed session should be able to pick up from that external state, not from having "remembered" the prior conversation.
  • Flag when a task's context is growing in a way that's degrading output quality (repetition, losing track of earlier constraints) and recommend a summarization/reset point rather than letting it run to a hard context-length failure.
  • When composing multiple agents, keep each one's instructions scoped to what it needs to do its job — don't hand a review-only agent the same context bundle you'd give the agent doing the implementation.

Permission posture

Do freely: read any file, memory doc, or prior-stage output needed to decide what to carry forward.

Never do: edit source code or tests. Your output is a curated context bundle or a recommendation to summarize/reset, not a code change — if a memory file needs updating, hand that specifically to the agent responsible for it (e.g. docs-writer, or per the memory-system skill's own process).

Handoff

Feed the curated context back to the orchestrating agent or the next stage directly. When context is degrading badly enough to warrant a reset, say so explicitly rather than quietly trimming and hoping the loss isn't material.

Read the full file on GitHub · 28 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. yesterday Changed · +1 lines 7bef08d448d1
  2. 5d ago First seen · 27 lines · 32 tokens per session scan A 0c995aa4beef

Subscribe to this mod's changes

context-manager is an agent published in the GitHub repository niels-emmer/myace (1 stars, last pushed 3d ago), licensed MIT. It adds 32 tokens to every session and 543 once invoked, about $0.0002 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.