Agent skills
Install CipherStash agent skills to give your AI coding assistant accurate knowledge of encryption setup, schema building, deployment, and integrations.
CipherStash Stack 1.0 publishes 13 agent skills covering the stable SDK, CLI, database surfaces, deployment workflow, and integrations. The skills give an AI coding assistant the exact APIs, commands, constraints, and rollout sequence it needs without making your application code part of the prompt.
Skills work with AI coding tools that support the skills protocol, including Claude Code, Codex, Cursor, GitHub Copilot, Windsurf, Cline, Gemini, AMP, Goose, Roo, and Trae.
Install every skill
Install the complete collection in your project root:
npx skills add cipherstash/stackThis installs all 13 skills. Your agent activates the relevant ones from the task and code it is working on.
Install through the Stash workflow
The stash 1.0 CLI bundles the skills and selects a smaller set for the integration detected by stash init. When stash impl hands the reviewed plan to an agent, it copies the selected skills into the tool's skill directory or includes their guidance in the generated agent instructions.
| Integration | Skills selected |
|---|---|
| Drizzle | stash-encryption, stash-drizzle, stash-indexing, stash-deployment, stash-zerokms, stash-auth, stash-cli |
| Supabase | stash-encryption, stash-supabase, stash-indexing, stash-postgres, stash-edge, stash-deployment, stash-zerokms, stash-auth, stash-cli |
| Prisma ORM 8 | stash-encryption, stash-prisma, stash-indexing, stash-deployment, stash-zerokms, stash-auth, stash-cli |
| PostgreSQL | stash-encryption, stash-indexing, stash-postgres, stash-edge, stash-deployment, stash-zerokms, stash-auth, stash-cli |
Claude Code receives skills under .claude/skills/; Codex receives them under .codex/skills/. Handoffs to tools that do not load a skill directory receive the same guidance through the generated agent instructions.
Available skills
| Skill | What it covers |
|---|---|
stash-encryption | The stable @cipherstash/stack EQL v3 typed schema, EncryptionClient, encryption and model operations, searchable fields, encrypted JSON, bulk operations, and migration lifecycle. |
stash-auth | Credentials, AutoStrategy, AccessKeyStrategy, OidcFederationStrategy, DeviceSessionStrategy, the four CS_* variables, lock contexts, and authentication failures. |
stash-zerokms | Keysets, clients, client keys, grants, revocation, key hierarchy, multi-tenant isolation, and diagnosing keyset mismatches. |
stash-indexing | Functional indexes over EQL v3 term extractors, managed-Postgres constraints, ORDER BY and GROUP BY shapes, and EXPLAIN verification. |
stash-deployment | The safe multi-deploy encryption rollout, credentials, backfill, read cutover, rollback, Prisma Compute, and preview environments. |
stash-cli | init, plan, impl, status, authentication, eql install/migration/repair/upgrade/status, validation, backfill, drop, schema generation, and the agent interface. |
stash-drizzle | @cipherstash/stack-drizzle 1.0 column factories, schema extraction, indexes, EQL v3 migration generation, and encrypted query operators such as eq, matches, contains, and selector. |
stash-supabase | The encryptedSupabase wrapper, encrypted mutations and reads, filters and ordering, identity-aware encryption, and PostgREST limitations. |
stash-prisma | @cipherstash/stack-prisma 1.0, encrypted Prisma column types, runtime envelopes, decryptAll, eql* query operators, and Prisma-owned EQL migrations. |
stash-postgres | Hand-written EQL v3 SQL with pg or postgres, typed bind parameters, operator selection, encryptQuery, and double-encoding failures. |
stash-edge | The @cipherstash/stack/wasm-inline entry for Deno, Supabase Edge Functions, Cloudflare Workers, and Bun, including explicit credentials and WASM-specific client behavior. |
stash-dynamodb | The @cipherstash/stack/dynamodb surface, encrypted attributes, HMAC query attributes, nested values, bulk operations, and DynamoDB table design. |
stash-supply-chain-security | Dependency admission, install cooldowns, lockfile integrity, frozen CI installs, registry pinning, Dependabot, CODEOWNERS, and npm provenance. |
How skills work
An installed skill activates when the task matches its description. Integration skills link back to the cross-cutting skills instead of duplicating their rules: authentication questions route to stash-auth, index design to stash-indexing, production rollout to stash-deployment, and key access to stash-zerokms.
That separation keeps the guidance consistent. A Drizzle, Supabase, Prisma, or raw PostgreSQL task uses the same released encryption model and deployment safety rules while adding only the query and schema behavior specific to that integration.
Typical workflow
- Initialize the project. Run
npx stash initwith--drizzle,--supabase, or--prismawhen appropriate. The CLI authenticates, inspects the project, installs compatible packages, and records context under.cipherstash/. - Draft a plan. Run
npx stash plan. The selected skills help the agent produce a reviewable.cipherstash/plan.mdwithout changing the application. - Review the plan. Confirm the selected columns, capabilities, indexes, migration sequence, and deployment gates.
- Implement it. Run
npx stash impl. The CLI hands the plan and integration-specific skill set to your chosen agent. - Track the rollout. Run
npx stash statusas schema changes, dual writes, backfill, read cutover, and plaintext removal progress.
Requirements
- Node.js 22 or later
- An AI coding tool that supports skills or generated agent instructions
- A CipherStash account (sign up)
Benchmarks
Query latency on encrypted columns versus plaintext PostgreSQL. Exact match and range run within 1.2-1.4x of plaintext and stay flat from 10k to 10M rows.
Glossary
Definitions of key CipherStash concepts and terms, from ZeroKMS, keysets, and client keys to EQL, HMAC, and searchable encryption.