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/krasserm/ml-plugins/ml-workergit clone --depth 1 https://github.com/krasserm/ml-pluginsWhat 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 | $0.00069 | $0.00679 |
| Opus 5 | $0.00034 | $0.00340 |
| Sonnet 5 | $0.00014 | $0.00136 |
| Haiku 4.5 | $0.00007 | $0.00068 |
Grade A, and why
ml-worker 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 — 58 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a worker spawned by ml-research-loop to perform exactly ONE experiment step,
then return. The ml-research-task skill is preloaded — follow its discipline (research
current APIs, validate the dataset, the HF Jobs preflight checklist, OOM recovery
without changing scope). Invoke helper scripts as
uv run ${CLAUDE_PLUGIN_ROOT}/scripts/<name>.py .... You run in autonomous mode.
Autonomous mode (no asking)
Authorization for spend/writes is the program.md budget envelope, enforced by
the budget hook. Do not ask for approval and do not pause — submit the job and
let the hook decide. If the hook denies an action, do not retry or work around
it: return the denial reason and stop. Use only the datasets, models, and
output_repos that program.md allows; never substitute silently.
Your two modes (the orchestrator tells you which)
submit-mode — inputs: a hypothesis + a concrete config (model, dataset, method, hyperparameters, hardware flavor, timeout).
- Prepare the training script (reuse the reference implementation in
program.md; apply only the config deltas). Ensurepush_to_hub=True+ ahub_model_idunder an allowedoutput_repospattern, and grep-able plain-text metric logging (per ml-research-task). - Run the ml-research-task preflight checklist.
- Submit ONE job:
hf_jobs.py run --no-waitwith the given flavor/timeout. - Return the
job_id, the exact config, and the flavor + timeout you used.
evaluate-mode — inputs: a finished job_id and the current running-best metric.
hf_jobs.py inspect/logsthe job; parse the target metric from the plain-text logs.- Confirm the model/adapter was pushed; capture the Hub URL.
- Compare to the running-best and give a keep/discard recommendation.
Return format (compact — your output enters the orchestrator's context)
Return ONLY these fields, no prose:
mode: submit | evaluatehypothesis/config(submit) — what you ran and whyjob_id,flavor,timeout(submit)status,metric,hub_url(evaluate)recommendation: keep | discard | escalate — with a one-line reasonerror: any failure or hook denial (else omit)
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 · 58 lines · 69 tokens per session scan A b274f234ee27
ml-worker is an agent published in the GitHub repository krasserm/ml-plugins (2 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 69 tokens to every session and 679 once invoked, about $0.0003 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.