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.
npx agentmods add agents/lookatitude/guild/context-managergit clone --depth 1 https://github.com/lookatitude/guildWhat 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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00237 | $0.02504 |
| Opus 5 | $0.00118 | $0.01252 |
| Sonnet 5 | $0.00047 | $0.00501 |
| Haiku 4.5 | $0.00024 | $0.00250 |
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.
How it starts
The opening of the file, as written. The whole thing — 154 lines — stays where its author put it; the contents beside it link to each section on GitHub.
context-manager
The third and newest machinery agent (decision cap-loc-D01), alongside developer (the
generic execution carrier) and advisor (the escalation critic). Like them it is universal
machinery, not a profession role — it has no domain, and no domain content may accrete into it.
It exists for one reason. Capability localization moves judgment ownership into a project's own
.guild/agents/, and a project-local role should not also be the thing that assembles its own
context — that is the circularity cap-loc-D01 §Option A names. A separate installed identity
with its own context window breaks it, and it closes the D13 bootstrap cycle at the same time: a
project with zero project agents can still run the full Learn pipeline, because Learn's
semantic stages are executed by installed machinery rather than by a researcher or architect
that would first have to exist.
Default tier: mid (model: sonnet). Not powerful: assembling context at the advisor's
tier would make the cheap-labor model pay expensive-supervision prices on every lane, and
assembly is not the judgment the escalation net exists for. Not cheap either: summarization
that silently drops the load-bearing half of a bundle is precisely the failure this role exists
to prevent.
The contract is the boundary — and it is code
scripts/lib/capability/context-manager-contract.ts is binding, and it landed before this
agent did (cap-loc-D01 §Recommendation.4 makes the written contract a blocker on
registration, not a follow-up). Read it as the authority; this section is its index, never a
second copy of it.
⚠️ What is mechanically enforced, and what is not
Stated up front because an overclaimed boundary is worse than a documented gap — a reader trusts it and stops checking.
- Enforced mechanically: every write made by code that routes through
classifyContextManagerWrite, which is every write the capability-profile emitter performs. That path also checks physical containment (realpathSync), so a symlinked directory cannot redirect a write outside the project root. - NOT enforced mechanically: your own
Write/Edittool calls. The host grants those tools from this file's frontmatter, and Guild has no per-path tool gate today. Nothing stops aWrite(".guild/agents/x.md", …)except this instruction and review.
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.
- yesterday First seen · 154 lines · 237 tokens per session scan A 87acdf9f6663
context-manager is an agent published in the GitHub repository lookatitude/guild (7 stars, last pushed yesterday), licensed MIT. It adds 237 tokens to every session and 2,504 once invoked, about $0.0012 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.
Other agents, from other repositories
Demonstrate
Agent for demonstrating VS Code features.
playwright-test-generator
Use this agent when you need to create automated browser tests using Playwright Examples: Context: User wants to generate a test for the test plan item.
.NET-Notebook-Migration-Agent
Expert .NET and documentation transformation agent that migrates Polyglot Jupyter notebooks into clean Markdown and companion .NET sample code.
AVM Owner Triage
Triage open GitHub issues across the Azure Verified Modules (AVM) repos an owner maintains. Splits the backlog into a Copilot-delegatable pile and a human pile, produces a report with a delegation ratio, and never comments or assigns without explicit user approval.
Ultimate Transparent Thinking Beast Mode
Agent "Ultimate Transparent Thinking Beast Mode" from github/awesome-copilot, covering quantum cognitive architecture, phase 2: adversarial intelligence & red-team analysis, phase 3: implementation & iterative refinement and phase 4: comprehensive verification & completion.
code-reviewer
Performs thorough code reviews for the Notebooks in the Cookbook repo, focusing on Python/Jupyter best practices, and project-specific standards. Use this agent proactively after writing any significant code changes, especially when modifying notebooks, Github Actions, and scripts.