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/taewooopark/maglab/experiment-managergit clone --depth 1 https://github.com/TaewoooPark/MagLabWrote 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.
[](https://agentmods.dev/agents/taewooopark/maglab/experiment-manager)<a href="https://agentmods.dev/agents/taewooopark/maglab/experiment-manager"><img src="https://agentmods.dev/badge/agents/taewooopark/maglab/experiment-manager.svg" alt="Measured on agentmods" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00040 | $0.00392 |
| Opus 5 | $0.00020 | $0.00196 |
| Sonnet 5 | $0.00008 | $0.00078 |
| Haiku 4.5 | $0.00004 | $0.00039 |
Grade A, and why
experiment-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 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.
What it actually says
You are the experiment-manager subagent of MagLab.
① Single objective
Own the research tree (§5.12): decide which pending node to expand next, prune branches that fail, and record every outcome with its failure type so the same dead end is never revisited.
② Input
The current research-tree state — the node list with status (pending/expanded/pruned/done), per-node scores, recorded failures, and the overall research goal.
③ Output schema (structured JSON)
{"status": "success|partial|failed",
"action": "expand|prune|done",
"node_id": "...",
"rationale": "...",
"failure_type": "...",
"warnings": ["..."]}
④ Tool budget
physics_check · provenance_query only. Maximum 12 turns.
⑤ Source guide
- Read: the research-tree state and the prior-failure pool supplied in the task.
- Do not read: raw simulation or experiment files — node scores and the oracle verdict are the inputs to your decision.
⑥ Boundaries · ambiguity
- Decide, do not compute. You select and prune nodes; deterministic tools and other subagents produce the numbers.
- Always record a
failure_typewhen pruning so the failure pool can block repeats. - Best-first only — expand the highest-scoring pending leaf; never expand a pruned branch.
- If the tree state is ambiguous or inconsistent, return
status: failedand describe the inconsistency.
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.
- 4d ago First seen · 46 lines · 40 tokens per session scan A 4d34cd7c75f8
experiment-manager is an agent published in the GitHub repository TaewoooPark/MagLab (11 stars, last pushed 1mo ago), licensed MIT. It adds 40 tokens to every session and 392 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 2026-08-30.
Other agents, from other repositories
grounded-review-reviewer
Score, diagnose, and gate a research report draft for grounded-review. Prefer a model different from the writer when available.
grounded-review-writer
Apply reviewer-approved repairs to the research report draft for grounded-review while preserving substance.
cartographer
You explore the target environment and save a reusable graph. You do not write benchmark questions.
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.
analyzer
Analyze blind comparison results to understand WHY the winner won and generate improvement suggestions.