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 StamKavid/last-ds-mile --skill ds-briefgit clone --depth 1 https://github.com/StamKavid/last-ds-mileWrote 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/stamkavid/last-ds-mile/ds-brief)<a href="https://agentmods.dev/skills/stamkavid/last-ds-mile/ds-brief"><img src="https://agentmods.dev/badge/skills/stamkavid/last-ds-mile/ds-brief/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/stamkavid/last-ds-mile/ds-brief"><img src="https://agentmods.dev/badge/skills/stamkavid/last-ds-mile/ds-brief.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 | $0.00089 | $0.01376 |
| Opus 5 | $0.00044 | $0.00688 |
| Sonnet 5 | $0.00018 | $0.00275 |
| Haiku 4.5 | $0.00009 | $0.00138 |
Grade A, and why
ds-brief 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 — 92 lines — stays where its author put it; the contents beside it link to each section on GitHub.
ds-brief — Executive Brief
Overview
/ds-report is written for a technical-enough stakeholder who can sit with RMSE,
AUC, fold spread, and a slice table. Most of the people who actually decide whether
a model ships are not that reader. This stage translates — it never re-derives —
/ds-report's findings into language a non-technical audience can act on without
first learning what a confusion matrix is.
When to Use
- After
/ds-reporthas produced its narrative. - Whenever asked to explain results to executives, a business audience, a board, or any stakeholder without a DS/ML background.
- NOT for: producing new findings or a different recommendation than
/ds-reportreached (that would be re-analysis, not translation) — if the technical report's conclusion needs to change, go fix/ds-reportfirst, then translate the corrected version.
Core Process
- Gate check: confirm
.last-ds-mile/stages/09-report.mdexists. If it doesn't, run/ds-report's work yourself now — this stage translates an existing narrative, it doesn't build one from raw evaluation results, but "existing" means "produced in this run if it wasn't already there," not "handed off to the user to go create." - Strip every metric name and statistical term — RMSE, AUC, PR-AUC, F-beta,
SHAP, p-value, standard deviation, fold, quintile/decile, calibration,
coefficient, feature importance, confound. If a sentence can't be said without
one of these words, translate it using
/ds-report's own cost-translation section (the dollar figure, rate, or count it already computed) — not by inventing a new plain-language number that wasn't in the technical report. - Structure as five short sections, in this order:
- What we built and why — one or two sentences, tied to the decision named in
/ds-frame. No architecture, no model names as a selling point. - How well it works — dollars, percentages, or counts only (e.g. "catches 9
of 10 at-risk customers" or "typical pricing error is about $20,000"), pulled
directly from
/ds-report's cost translation. - Where it's weaker — named plainly with a real-world reason, not a statistical one (e.g. "less reliable for the cheapest homes, which include more distressed or family sales" — not "higher RMSE in the Q1 slice").
- The recommendation — stated as plainly as
/ds-report's own, e.g. "deploy as a suggestion, with a person reviewing the flagged cases." - What happens next — who owns the decision, and the one or two assumptions
that would need to hold (from
/ds-report's Assumptions section) for the recommendation to still be right.
- What we built and why — one or two sentences, tied to the decision named in
- No new claims. Every sentence must trace back to something already stated in
/ds-evaluate,/ds-explain, or/ds-report. This stage's job is simplification, not additional analysis — if a genuinely new finding surfaces while writing this, that belongs back in/ds-report, not smuggled into the brief alone. - Length discipline: roughly 300–500 words, one page. A "brief" the length of the technical report has failed at the one thing it exists to do.
- Write to
.last-ds-mile/stages/09b-brief.md— a companion to09-report.md, not a renumbering of the pipeline; not every project needs this artifact, so it isn't a gate anything downstream depends on.
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 · 92 lines · 89 tokens per session scan A a5279f320f00
ds-brief is a skill published in the GitHub repository StamKavid/last-ds-mile (3 stars, last pushed 1mo ago), licensed MIT. It adds 89 tokens to every session and 1,376 once invoked, about $0.0004 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
developing-with-streamlit
Use for ALL Streamlit tasks: creating, editing, debugging, beautifying, styling, theming, optimizing, or deploying Streamlit apps. Also custom components, st.components.v2, HTML/JS/CSS work. Discovers and loads version-matched reference docs from the user's installed Streamlit (>=1.57). Triggers: streamlit, st.…
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.
wax
Swift framework guidance for Wax on-device memory/RAG. Use when writing Swift code with the public Memory facade, experimental PhotoMemory / VideoMemory, BuiltInMultimodalEmbeddings, embedding providers, retrieval modes, or hybrid search. For agent operators using the Wax MCP server tools, use the separate wax-mcp…
wax-memory-maintenance
Maintain the Wax memory store used by the Hermes wax-memory provider.
ml-expert
Expert-level machine learning, deep learning, model training, and MLOps. Use when the user mentions machine learning, deep learning, neural networks, MLOps, or data science, or when the task involves Machine Learning Fundamentals, Data Preparation, or Model Training.
fast-dash
Build a Fast Dash web app from a Python function. Use when the user wants to turn a function into an interactive app, add a UI to an existing function, or build a dashboard / form / wizard. Fast Dash infers UI components from type hints, so a well-typed function becomes an app with one decorator.