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 skills add pengzhangzhi/superpowers-ml --skill ml-result-reviewgit clone --depth 1 https://github.com/pengzhangzhi/superpowers-mlWrote 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/skills/pengzhangzhi/superpowers-ml/ml-result-review)<a href="https://agentmods.dev/skills/pengzhangzhi/superpowers-ml/ml-result-review"><img src="https://agentmods.dev/badge/skills/pengzhangzhi/superpowers-ml/ml-result-review/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/pengzhangzhi/superpowers-ml/ml-result-review"><img src="https://agentmods.dev/badge/skills/pengzhangzhi/superpowers-ml/ml-result-review.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00029 | $0.00845 |
| Opus 5 | $0.00015 | $0.00423 |
| Sonnet 5 | $0.00006 | $0.00169 |
| Haiku 4.5 | $0.00003 | $0.00085 |
Grade A, and why
ml-result-review 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 12d 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.
How it starts
The opening of the file, as written. The whole thing — 81 lines — stays where its author put it; the contents beside it link to each section on GitHub.
ML Result Review
Overview
You are concluding an ML experiment, not selling it. Passing tests showed the code path runs; they never showed the method works. Your job here is a SOBER conclusion that resists over-claiming, written so your human partner can decide on real evidence.
Core principle: Confirmed, exploratory, and failed are different things. Keep them apart.
This is downstream of superpowers-ml:ml-feedback-ladder (which owns the rungs) and superpowers-ml:verification-before-completion (evidence before claims). Cite the highest rung you actually reached, never the one you hoped for.
What "Confirmed" Means
A result is CONFIRMED only when the required full or equivalent evaluation completed under the declared primary metric and met the locked success criterion.
- Met the locked criterion under the locked metric = confirmed.
- Better-looking but not the locked test = exploratory, not confirmed.
- Operational interruptions (preemption, requeue, node failure) are not scientific failures. A requeued run that still completed the locked evaluation is confirmed. Note the interruption; do not downgrade the result for it.
Never claim the method beats a baseline without the full or equivalent evaluation under the locked primary metric.
The Review Format
Write the conclusion under these explicit headings, in order:
CONFIRMED
What met the locked success criterion under the primary metric. One line per claim, each tied to the evaluation that proved it. If nothing is confirmed, write "None."
EXPLORATORY
Interesting observations that were NOT the locked test: secondary metrics, partial runs, eyeballed curves, single seeds. Label clearly so no one mistakes these for results.
FAILED / INCOMPLETE
Runs that did not finish the locked evaluation, and WHY. Distinguish scientific failure (method/config wrong) from operational interruption (infra). An operational interruption belongs here only if it actually prevented the evaluation from completing.
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.
- 12d ago First seen · 81 lines · 29 tokens per session scan A 565eeb647e55
ml-result-review is a skill published in the GitHub repository pengzhangzhi/superpowers-ml (8 stars, last pushed 5d ago), licensed MIT. It adds 29 tokens to every session and 845 once invoked, about $0.0001 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 skills, from other repositories
hugging-face-model-trainer
This skill should be used when users want to train or fine-tune language models using TRL (Transformer Reinforcement Learning) on Hugging Face Jobs infrastructure. Covers SFT, DPO, GRPO and reward modeling training methods, plus GGUF conversion for local deployment. Includes guidance on the TRL Jobs package, UV…
hugging-face-datasets
Create and manage datasets on Hugging Face Hub. Supports initializing repos, defining configs/system prompts, streaming row updates, and SQL-based dataset querying/transformation. Designed to work alongside HF MCP server for comprehensive dataset workflows.
grpo-finetune
Fine-tune a model with GRPO on Fireworks-managed GPUs from a plain-English task description and a dataset. Use this skill whenever the user wants to fine-tune, RL-tune, or GRPO-train a model on their own data — or says things like "train a model to extract/classify/score X", "fine-tune on this dataset", "set up a GRPO…
marimo-pair
Work inside the user's live marimo notebook from the code editor: run Python in the same kernel the user does, inspect live notebook state, and commit durable notebook changes through code mode. Use whenever you create, analyze, or improve the user's marimo notebook.
training-check
Interactively monitor training metrics from the current Codex session, periodically checking WandB or fallback logs for NaN, divergence, plateaus, and broken runs.
pysr
Use when fitting equations to data with PySR or SymbolicRegression.jl, when a user wants an interpretable formula, symbolic model, scaling law, or empirical relation discovered from numeric data, or when debugging a PySR search that is slow, stuck, or giving poor equations.