system2
system2 contains 17 collected skills from diegoscarabelli, with repository-level occupation coverage and site-owned skill detail pages.
Skills in this repository
Use when editing `~/.system2/config.toml` — adding or modifying a database connection, tuning per-agent overrides, or changing operational defaults. Does NOT cover `~/.system2/auth/.auth.toml` (managed by `system2 config`).
Run on first launch (when ~/.system2/knowledge/infrastructure.md is still the template) or whenever the user explicitly asks to re-onboard. Greets the user, learns about them, detects the system, installs and configures the data stack (database, orchestrator, Python environments), scaffolds the pipeline repository, and captures interaction preferences.
Use when building, debugging, deploying, or testing data pipelines with Prefect v3. Trigger on any code importing prefect, prefect YAML config, or user mentioning Prefect flows/tasks/deployments/workers.
Run before declaring project work complete. A self-audit checklist to honestly assess whether the deliverables actually meet the requirements before requesting final review or reporting completion to the Guide.
Run when delegating complex work to a new project. Gathers preliminary requirements with the user, creates the project in app.db, spawns a Conductor and Reviewer, introduces them to each other, updates the user, and schedules a follow-up reminder so a silent Conductor is noticed.
Build interactive HTML dashboard artifacts with live database queries via the postMessage bridge. Includes a copy-paste helper function, workflow steps, and Chart.js integration patterns.
Use when building, debugging, scheduling, or testing data pipelines with Apache Airflow v3. Trigger on any code importing airflow, DAG definitions, operator usage, or user mentioning DAGs/tasks/operators/sensors/scheduling.
Run when the Conductor reports that project work is complete. Confirms with the user, tells the Conductor to close the project, waits for the close-project report, then terminates the Conductor and Reviewer and marks the project done.
Conventions for building Python ETL/ELT pipelines: file state machine (ingest/process/store/quarantine), standard task sequence with customization hooks, config-first parametrization, per-pipeline directory layout, SQLAlchemy ORM integration (make_base, fkey, upsert). Framework-agnostic; applies to Airflow, Prefect, and others.
Use when reviewing code for correctness, security, design, performance, SQL quality, testing, and readability. Trigger on any code review request, PR review, or pre-push validation.
Column-level schema details for all seven app.db tables. Read this when you need exact column names, types, constraints, or indexes before writing queries or creating/updating records.
Run when the user wants to revisit or continue work on a completed project. Helps the user weigh resurrection against a new project or a bespoke task, then resurrects the original Conductor and Reviewer with their context intact and reopens the project record.
Use when reviewing analytical work for cognitive biases and reasoning fallacies. Trigger on any review of conclusions, causal claims, risk assessments, forecasts, or interpretive analysis where System 1 thinking may have gone unchecked.
Use when designing database schemas, choosing between normalization levels, modeling dimensional/star schemas, deciding on JSON columns vs relational columns, choosing materialization strategies (materialized views vs pipeline-built tables), or reviewing table structures. Trigger on CREATE TABLE statements, schema design discussions, materialization decisions, or questions about data modeling patterns.
Use when performing or reviewing quantitative analysis, choosing statistical tests, checking assumptions, interpreting results, or conducting Bayesian modeling. Trigger on any work involving hypothesis testing, effect sizes, confidence/credible intervals, regression, or statistical reporting.
Use when designing schemas, writing queries, configuring compression/retention/continuous aggregates, or troubleshooting performance for TimescaleDB hypertables. Trigger on any SQL referencing hypertables, time_bucket, continuous aggregates, or TimescaleDB functions.
System2 UI layout and panel reference. Read this when the user asks about the interface, when you need to direct them to a specific panel, or when you need to describe what they are seeing.