Cortex Code Skills

This is a curated collection of Agent Skills for Cortex Code ("CoCo") — Snowflake's CLI for building with AI.

Each skill is a folder containing a SKILL.md that teaches CoCo a workflow, coding standard, or domain-specific best practice. CoCo already ships with 50+ bundled skills (run /skill to browse). The skills here extend that catalog with workflows contributed by Data Superheroes, Snowflake employees, and partners.

• Install these skills
• Skill catalog
• Authoring a skill
• Repo structure
• Troubleshooting
• Disclaimer

---

Install these skills

Open Cortex Code and ask:

Install the skills from https://github.com/Snowflake-Labs/cortex-code-skills

CoCo will clone, cache, and register every skill in this repo. To pull updates later:

Sync my skills

To install a single skill rather than the whole repo:

Install the skill at https://github.com/Snowflake-Labs/cortex-code-skills/tree/main/skills/<skill-name>

Once installed, invoke a skill by typing $<skill-name> followed by your prompt. Run /skill list to see what's loaded, or $$ to verify the active session's skill set.

Snowflake connection required. Most skills here run SQL or call Cortex services. Set your active connection with cortex connections set <name> before invoking.

---

Skill catalog

Snowflake docs & learning

Skill	What it does
snowflake-docs	Answer any Snowflake question by searching official docs via the Cortex Knowledge Extension.
quickstart-guide	Paste a Snowflake Quickstart URL and get a guided, interactive walkthrough.

Analytics & semantic modeling

Skill	What it does
semantic-view-patterns	Apply 25 production-tested Semantic View patterns covering joins, metrics, dimensions, and access policies.
ontology-stack-builder	Build a 5-layer Ontology-on-Snowflake stack (physical → metadata → abstract views → semantic views → Cortex Agent) from a relational schema or OWL file.

AI & Cortex

Skill	What it does
entity-resolution	End-to-end entity resolution pipeline using Cortex AI Functions to match, link, and dedupe records.

Data engineering & integration

Skill	What it does
snowpipe-bcdr	Snowpipe disaster-recovery patterns for Azure ADLS Gen2 — failover, failback, and catchup.
openflow-spcs-privatelink	Set up AWS PrivateLink between OpenFlow on SPCS and private sources like RDS or on-prem databases.
manage-zerocopy-sapbdc	Manage the SAP Business Data Cloud zero-copy connector lifecycle: create, enroll, consume, publish, analyze.

Operations, MLOps & governance

Skill	What it does
rbac	Design Snowflake RBAC hierarchies and access-role patterns.
mlops	Router skill for MLOps on Snowflake — maturity assessment, promotion patterns, CI/CD, monitoring, governance.
dcr-v1-to-v2	Migrate a Data Clean Room from the V1 SAMOOHA Provider/Consumer API to the V2 Collaboration API.

---

Authoring a skill

1. Start from an existing skill — pick any skill in skills/ whose shape matches what you're building (e.g. rbac for a router skill, entity-resolution for a multi-step pipeline). Copy its folder and rename to your skill's name.
2. Fill in the frontmatter — name, title, summary, description with triggers, type (community | snowflake | partner), and tools you'll use.
3. Write the body — Overview, Workflow with numbered steps, Common Mistakes. Aim for under ~500 words; move reference material to references/<topic>.md, executable helpers to scripts/.
4. Pick a license — Apache 2.0 for community contributors, Snowflake license for employees. The license file lives inside the skill folder, not at the repo root.
5. Test in CoCo — load your local skill with Install the skill at <path> and run a few sessions against representative tasks.
6. Open a PR — see CONTRIBUTING.md for the review checklist.

---

Repo structure

skills/
  your-skill-name/
    SKILL.md                   # required
    LICENSE                    # required (Apache 2.0 or Snowflake license)
    references/                # optional — additional docs loaded on demand
      patterns.md
    scripts/                   # optional — executable helpers (Python, bash)
      validate.py
    assets/                    # optional — templates, fixtures, sample data
      example.csv

references/ and scripts/ are first-class. Use them to keep SKILL.md focused on the workflow itself; CoCo loads supporting files only when the workflow points at them.

---

Troubleshooting

Problem	Fix
/skill list doesn't show a skill from this repo	Re-run the install prompt above, then run /skill sync.
Skill triggers in unexpected sessions	Tighten the description field — keep triggers specific, list "Do NOT use for…" cases for adjacent skills.
Skill never triggers	Trigger keywords may be too narrow or too generic. Add concrete user phrases the skill should respond to.
Name collision with a bundled skill	A repo skill silently shadows a bundled skill of the same name. Run /skill list to spot duplicates and rename if needed.
SQL fails immediately	Most skills need an active Snowflake connection. Run cortex connections list and cortex connections set <name>.

For deeper debugging, read docs.snowflake.com/.../cortex-code/extensibility.

---

Disclaimer

These skills are contributed by the community for educational and reference use. Snowflake reviews PRs before merging but cannot guarantee correctness, completeness, or security. Review each skill's SKILL.md, LICENSE, and any bundled scripts/ before loading — especially skills that execute SQL, modify Snowflake objects, or call external services.

Snowflake reserves the right to remove skills that violate our contribution guidelines or terms of use.