High-Stakes Analytics Lab skill
High-Stakes Analytics & Decision Lab
An evidence-gated analytics skill for high-stakes decisions: profiles data, picks a method the evidence supports, can answer with a data request
- 1,008 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Profiles and analyzes the data file you hand it, which may contain personal or commercial information
- Runs Python scripts locally on your machine for calculations and chart generation
- Writes report, data-profile and chart files to disk in the folder you point it to
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/high-stakes-analytics-decision-labDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Profiles and analyzes the data file you hand it, which may contain personal or commercial information
- Runs Python scripts locally on your machine for calculations and chart generation
- Writes report, data-profile and chart files to disk in the folder you point it to
This is third-party code. Review the repository files before installing.
What it does
This agent skill works through data behind consequential decisions in a strict order: state the question, population, unit of observation, target quantity and horizon first, then check the data's readiness, and only then move to analysis. The `python3 scripts/hsadl.py doctor` command checks the environment before a run, and `python3 scripts/hsadl.py demo` sets up a tutorial example on synthetic data the skill explicitly forbids citing as real evidence. The `data_quality.py` script profiles the source file, preserves the original unchanged and pauses on material data transformations, while `analytics_router.py` uses that profile to decide which of four analysis types (descriptive, diagnostic, predictive, prescriptive) to add, without padding the report with ones the evidence doesn't call for. Diagnostic and prescriptive work separately tracks shared sources of uncertainty — time, market, participants, campaigns, operations and spatial proximity — so dependent risks aren't treated as independent ones. The finished package keeps a report.md narrative, a machine-readable results.json, a figures/chart-map.json that ties every chart to its question and source, and traceable results down to file hashes, together in one place. The skill treats more than a recommendation as an honest outcome: a request for more data, a negative model-validation result or a flat 'do not deploy' verdict count as a legitimate ending, same as a bounded action.
Who it is for. For analysts and product managers who need to work through real data behind a high-stakes decision and get a report that honestly stops where the data falls short of a conclusion.
Good fit when
- You need to work through a data export behind a question with real consequences and don't want a forced recommendation
- You need to profile an unfamiliar file before analysis: quality, missing values, personal data
- You need a predictive or prescriptive model with an explicit account of dependent uncertainty, not a formal confidence interval alone
- You want a report where every figure and chart traces back to the source file and its hash
Not a fit when
- You want a quick draft chart with no data profiling or evidence trail: the skill is built for high-stakes decisions, not a one-off visualization
- There is no data for the question at all: the skill will not invent an analysis from nothing, it honestly returns a request for more data
- You want a fixed-section report generator: the router only adds the analysis types the data profile actually supports
Example request
Run the readiness check on this sales data file, preserve the original and pick only the analysis types the data actually supportsLimitations
License is MIT.
- The skills installer connects only the compact `skills/high-stakes-analytics-decision-lab/` package, 39 files and about 472 KiB, not the whole repository with its demo portal and case archive.
- The example from the `demo` command is built on synthetic engineering data the skill explicitly forbids citing as real evidence.
- The profiling script pauses on material transformations of the source data and waits for your decision, so a fully unattended run is not supported.
- The scripts run locally on the files you provide and have no built-in calls to external APIs or cloud storage.
- A machine-readable file and hash contract lives in `bundle-manifest.json`, which lets you verify nothing was lost when the skill is updated.
How to disable. Delete the `high-stakes-analytics-decision-lab` folder from the agent's personal skills folder. Report and data-profile files you pointed to with --output-dir stay on disk and are removed separately.
FAQ
Can the skill produce a report with no recommendation?
Yes, that's a built-in honest outcome: a request for more data, a negative model-validation result or a 'do not deploy' verdict count as a legitimate ending, same as a bounded action. The report template does not push a recommendation when the evidence falls short.
What does the doctor command check before work starts?
It checks the environment: the required Python packages and script availability, before the agent moves to an executable analysis. The README names it as the required first step before any workflow.
What happens to the source data file during the quality check?
The data_quality.py script preserves the original unchanged and pauses when a transformation materially changes its meaning, waiting for a human decision. The skill does not auto-correct data without your involvement.
How does the skill handle dependent uncertainty?
In diagnostic and prescriptive analysis it separately tracks shared risk sources — time, market, participants, campaigns, operations, spatial proximity — so correlated risks aren't treated as independent the way a plain confidence interval would.
Can I trust the report's charts without checking their source?
Each chart is described in chart-map.json, tying it to a specific question and data source. The full results package includes file hashes you can check to confirm a chart was built on exactly the export that went in.
Discussion
No questions or reviews yet.
Related
SkillsA data visualization skill for agents: turns data into polished interactive HTML charts and full reports with one visual language
SciPilot Figure
scipilot-figure-skill
A scientific visualization advisor skill: it profiles the data and picks a chart for the argument first, then renders it to journal-grade quality
A skill and Python package for hyperparameter tuning: the agent reads the code, proposes the next set of values, and your own run measures the result
A library of agent skills for medical research: literature search and appraisal, study design, data analysis and academic writing
166 skills for scientific work: bioinformatics, cheminformatics, clinical data, geospatial analysis and 100+ databases
A three-stage workflow for math modeling tasks: analysis and model choice, Python or MATLAB computation, and paper assembly in Word or LaTeX
README in short
The repository description opens with a short explanation of why it exists: high-stakes analysis often fails before the model, through a vague question and hidden data-cleaning decisions. A six-principle table with the system's behavior under each follows, plus a three-step quick start. A separate section gathers fifteen complete example analyses across different fields, linking to a live example on the project's site.
README badge
Original SKILL.md
--- name: high-stakes-analytics-decision-lab description: Build or review source-backed descriptive, diagnostic, predictive, and prescriptive analysis for consequential decisions. Use when an agent must profile and safely prepare uploaded data, turn a real dataset or research question into a reproducible study, investigate drivers without overstating causality, validate a model, compare feasible actions under dependent uncertainty and tail risk, trace every parameter to evidence and approval, or produce an answer-first analytical report across health, business, finance, policy, engineering, operations, behavioral science, AI, or planning. --- # High-Stakes Analytics & Decision Lab Turn a real question into a defensible path from source evidence to analysis and, only when justified, bounded action. Keep observation, diagnosis, prediction, causal evidence, value judgments, and recommendation visibly separate. ## Start here Run the environment audit before an executable workflow: ```bash python3 scripts/hsadl.py doctor ```