Engineering and documentation skills maintained by NSX MarTech.
The canonical contract is /.agents/install-block.md. Quick version:
git clone https://github.com/NSXBet/martech-skills.git ~/martech-skills
~/martech-skills/install.sh
# then restart your harnessUpdate:
git -C ~/martech-skills pull --ff-only
~/martech-skills/install.sh| Skill | What it does |
|---|---|
martech-doc |
MarTech Technical Review & Approval — 9-section walk + anti-slop + Notion paste shape |
martech-medallion |
Engineering advice and readiness reviews for ingestion, Bronze/Silver/Gold, data quality, history, and Databricks governance |
martech-visualizations |
Sourced dashboard knowledge — schema/joins, KPI definitions, conventions, and SQL for the eight MarTech dashboards (FTD Tier, Payback, Copa, LTV, Attribution, Last Touch, Cockpit, Paid Media) |
New skills go in .agents/skills/<kebab-case-name>/. See TEMPLATE.md.
| Harness | Root |
|---|---|
| Claude Code | ~/.claude/skills/ |
| Codex CLI / generic | ~/.agents/skills/ |
| OMP (Oh My Pi) | ~/.agents/skills/ (Agents discovery; supported by inspected OMP 18.1.14) |
| Cursor | ~/.cursor/skills/ and ~/.agents/skills/ |
| OpenCode | ~/.config/opencode/skills/ |
| Gemini/Goose/Amp/Copilot | ~/.agents/skills/ (untested) |
After installing/updating and restarting your harness, ask it to use martech-medallion. For example:
Use martech-medallion to advise on this ad-platform ingestion design.
Define Bronze/Silver/Gold responsibilities, the data contract, and validation checks.
Use martech-medallion to review these pipeline definitions and DQ reports.
Assess readiness for the specified brands, history, metrics, and consumers.
In Codex, invoke $martech-medallion with the request. Natural-language invocation keeps the examples portable across harnesses. OMP must have Agents skill discovery enabled; higher-priority copies with the same name can shadow this version.
The skill includes layer standards, readiness evidence formats, a contract/dictionary guide, and dated research and decisions. It proposes a team baseline and checks actual evidence; Data Platform retains authority over local governance. It can work from supplied artifacts without a Databricks connection and labels unexecuted checks accordingly.
- Branch, add
.agents/skills/<name>/withSKILL.md(fromTEMPLATE.md). bash scripts/validate.shandfor f in install.sh scripts/*.sh; do bash -n "$f"; done.- Open a PR. Do not include production data, PII, credentials, tokens, connection strings, or confidential internal identifiers.