Investigation workflows for AI agents exploring geometric data spaces.
Structured behaviors that guide AI agents through exploration, investigation, detection, and monitoring of geometric data spaces. Not code — decision frameworks with examples.
| Skill | Focus | Typical trigger |
|---|---|---|
gds-analyst |
Investigation driver — goal-oriented, reads hints, plans approach | "investigate this sphere", "find anomalies" |
gds-explorer |
Orientation — understand a sphere you've never seen | "explore this sphere", "give me an overview" |
gds-investigator |
Root cause tracing, entity deep-dive, hypothesis testing | "why is this anomalous", "validate this finding" |
gds-detective |
Detection cookbook — event rates, composites, drift, passive_scan | "detect copula", "windowed comparison" |
gds-scanner |
Advanced detection — cross-pattern, neighbor, trajectory, segment shift | "cross-pattern discrepancy", "trajectory shape" |
gds-monitor |
Monitoring — drift detection, regime changes, temporal health | "check sphere health", "detect drift" |
gds-fraud-investigator |
AML/fraud investigation and 25 typology recipes | "detect fraud", "run AML scan" |
gds-sphere-designer |
End-to-end sphere design, construction, and tuning | "design a sphere", "build from my data" |
Design: gds-sphere-designer
Investigate: gds-analyst (drives the investigation)
├── gds-explorer (orientation, profiling)
├── gds-investigator (root cause, entity 360)
├── gds-detective (detection recipes)
├── gds-scanner (advanced detection)
├── gds-monitor (temporal, drift, regimes)
└── gds-fraud-investigator (AML/fraud domain)
gds-analyst drives investigations — reads the task, picks the right approach, delegates to specialized skills. The other skills are the toolbox.
npx skills add hypertopos/hypertopos-skillsWorks with Claude Code, Cursor, Codex, GitHub Copilot, Amp, Cline, Windsurf, and other supported agents.
Copy the gds-*/ skill folders into your agent's skills directory (e.g. .claude/skills/ for Claude Code).
- hypertopos-mcp available and connected
- hypertopos installed in the same environment
- A built sphere for skills that operate on live data
Skills are intentionally reactive — decision frameworks, not rigid checklists. MCP provides the tools and sphere data. Skills provide judgment, scope, and workflows.
Research-stage project. Skills and workflows are actively evolving.