25 Cursor Slash Commands and Rules Combos Power Users Swear By in 2026

By Rakshit Yadav (@yadavrakshit60)•Oct 2026•12 min read

Short answer: a slash command alone only fires when you remember to type it; a rule alone only knows what's in the file, not what you're about to do. Pairing one .cursor/commands/*.md file with one scoped .cursor/rules/*.mdc file closes both gaps — the rule keeps the guardrail always loaded for matching files, the command gives you an on-demand action that assumes the rule already ran. Below are 25 combos for a solo SaaS stack: 10 for billing, auth and CRO, 10 for testing, CI and docs, and 5 advanced chains where one command's output feeds the next.

If you haven't read the single-primitive posts yet, start with 12 copy-paste Cursor commands and 10 copy-paste Cursor rules. This post assumes you know the basics and focuses on the pairing pattern itself — which is a different problem than either post solves alone.

Why commands alone aren't enough

A command is a saved prompt you invoke by typing /name. It has no memory of your stack between invocations beyond what's in its own text, and it only runs when you type it. That's fine for one-off actions like /pr-description, but it breaks down for anything the agent should already be doing silently while you work the command — like "never touch a migration without a down path" or "every webhook handler verifies a signature." If that constraint only lives inside the command's prompt text, it does nothing the other 95% of the time you're editing those files without typing the command.

A rule is the opposite problem. alwaysApply: true or a globs match loads the rule automatically, but a rule is passive — it shapes how the agent writes code, it doesn't do anything. You can't "run" a rule the way you run a command.

The fix is to stop treating them as competing primitives and start pairing them: the rule is the standing guardrail that's always in context for the files it covers, the command is the explicit action you trigger that can safely assume the guardrail already applied. Write the rule first, then write the command's prompt to reference what the rule guarantees instead of re-stating it — that's the whole pattern, and it's what turns "I have 40 commands and 20 rules" into something more than a pile of unrelated files.

10 combos for SaaS billing, auth, and CRO

Each pair: the rule that's always loaded for matching files, then the command you run on top of it.

1. Stripe webhook idempotency

---
description: Stripe webhook handlers must verify signatures and dedupe by event ID
globs: app/api/webhooks/stripe/**/*.ts
alwaysApply: false
---
- Verify the Stripe signature with the raw body before parsing JSON.
- Check event.id against a processed_events table before acting; insert it in the same transaction as the side effect.
- Never trust amounts or status from the payload alone — refetch the object from Stripe's API inside the handler.
/webhook-audit
Audit the Stripe webhook handlers I point at against our idempotency rule.
Flag any handler missing a signature check, missing dedupe, or trusting payload data without a refetch.
Fix the gaps in place; don't touch handlers that already pass.

2. Billing state machine

---
description: Subscription status transitions must go through the state machine, not direct writes
globs: lib/billing/**/*.ts
alwaysApply: false
---
- Never write `subscription.status` directly. Call the transition function for the target state.
- Every transition logs the from-state, to-state, and triggering event.
- Dunning transitions (past_due, canceled) must check for an active retry campaign before firing.
/billing-fix
Find every place in lib/billing that writes subscription status.
Replace direct writes with the correct transition function. If no transition function covers the case, tell me before adding one.

3. Auth on server actions

---
description: Every server action checks the session and tenant scope before touching data
globs: app/**/actions.ts,app/**/*actions*.ts
alwaysApply: false
---
- First line of every exported action: resolve the session, return early if absent.
- Every query filters by the resolved org/tenant ID — never trust an ID passed from the client.
- Mutations re-check ownership of the row being modified, not just that a session exists.
/security-pass
Check every server action in the files I point at against the auth rule.
Report any action missing a session check, missing tenant scoping, or trusting a client-supplied ID for ownership.

4. Rate limiting on auth endpoints

---
description: Auth endpoints (login, signup, password reset) must be rate-limited
globs: app/api/auth/**/*.ts
alwaysApply: false
---
- Every endpoint here calls the rate limiter before touching the database.
- Limit by IP and by the submitted identifier (email/username), not just one or the other.
- Return the same response shape on rate-limit and on invalid-credentials so timing doesn't leak which one failed.
/rate-limit-check
Check the auth endpoints I point at against the rate-limit rule.
Add the missing limiter calls using our existing rate-limit helper. Don't invent a new limiter.

5. Pricing page copy

---
description: Pricing page copy must match the actual plan limits in the billing config
globs: app/(marketing)/pricing/**/*.tsx
alwaysApply: false
---
- Never hardcode a plan price or limit in JSX. Pull it from the billing config file.
- Every claim ("unlimited X", "up to N seats") must be checkable against that config.
/pricing-audit
Compare the pricing page copy against lib/billing/plans.ts.
List any claim that doesn't match the config, and fix the copy to match — don't change the config.

6. Checkout CRO guardrail

---
description: Checkout flow changes must not add a field or step without justification
globs: app/checkout/**/*.tsx
alwaysApply: false
---
- Count the form fields and steps before and after any change. A net increase needs an explicit reason in the PR description.
- Error states must be inline, not a redirect to a generic error page.
/checkout-review
Review my changes to the checkout flow against the CRO rule.
If fields or steps increased, ask me for the justification instead of assuming one. Check error states are inline.

7. Email/notification copy

---
description: Transactional emails must include a working unsubscribe/preference link and match the brand voice doc
globs: lib/email/templates/**/*.tsx
alwaysApply: false
---
- Every template includes the preference-center link from lib/email/footer.tsx.
- Subject lines are under 60 characters and avoid ALL CAPS and excessive punctuation.
/email-audit
Check every template in lib/email/templates against the rule.
Flag missing footer includes and subject lines over 60 characters. Fix in place.

8. Refund and chargeback handling

---
description: Refunds and chargebacks must update entitlements, not just the Stripe record
globs: lib/billing/refunds/**/*.ts
alwaysApply: false
---
- A refund that drops the customer below their plan's usage threshold must revoke the matching entitlement in the same transaction.
- Chargebacks always flag the account for manual review; never auto-reinstate.
/refund-check
Audit the refund handling code against the rule.
Report any path where a refund or chargeback doesn't touch entitlements, and fix it.

9. Feature-flag cleanup

---
description: Feature flags older than 90 days must be resolved, not left in code
globs: lib/flags/**/*.ts
alwaysApply: false
---
- Every flag definition includes a created date.
- A flag past 90 days with no removal PR linked is a code smell, not a permanent fixture.
/flag-cleanup
List every flag in lib/flags older than 90 days per the rule.
For each, tell me whether the code path can be deleted (flag always on/off in practice) or still genuinely split.

10. Onboarding funnel instrumentation

---
description: Every onboarding step must fire an analytics event with a consistent step name
globs: app/onboarding/**/*.tsx
alwaysApply: false
---
- Each step component fires `track("onboarding_step", { step: "<name>" })` on mount and on completion.
- Step names must match the list in lib/analytics/onboarding-steps.ts — no ad hoc strings.
/onboarding-audit
Check onboarding step components against the instrumentation rule.
Flag missing or mismatched event names and fix them to match lib/analytics/onboarding-steps.ts.

10 combos for testing, CI, and docs

11. No skipped tests merge silently

---
description: Skipped or TODO tests must have a linked issue in the comment
globs: "**/*.test.ts,**/*.spec.ts"
alwaysApply: false
---
- `it.skip` or `describe.skip` must have a `// TODO(issue-url):` comment on the line above.
- A skip without that comment is a bug, not a style nit.
/test-audit
Find every skipped test in the files I point at.
Flag any skip missing the issue-link comment per the rule. Don't un-skip anything — just report.

12. Test file mirrors source structure

---
description: Test files live next to the source file they test, not in a separate __tests__ tree
globs: "**/*.test.ts"
alwaysApply: false
---
- `foo.ts` is tested by `foo.test.ts` in the same directory.
- Shared test helpers go in `test-utils/`, not duplicated per file.
/write-tests
Write tests for the file I point at, placed per the rule (same directory, matching name).
Reuse existing helpers from test-utils/ instead of writing new mocks.

13. CI must stay green before merge

---
description: No PR merges with a red check; flaky tests get quarantined, not ignored
globs: .github/workflows/**/*.yml
alwaysApply: true
---
- A failing check blocks merge. There is no "merge anyway" exception documented here.
- A test that fails intermittently gets moved to a quarantine job with a tracking issue, not deleted.
/ci-triage
Look at the current failing CI run.
Classify each failure: real regression, flaky (quarantine per the rule), or environment issue.
For real regressions, point to the file:line cause.

14. Migration safety gate

---
description: Migrations must be reversible and safe to run against a table with existing rows
globs: "migrations/**/*.sql,db/migrate/**/*.ts"
alwaysApply: false
---
- New required columns are added nullable or defaulted first; a separate migration adds the constraint after backfill.
- Every migration file has a working down/rollback.
- Note any lock or long-running risk in a comment at the top of the file.
/db-migration
Write the migration for the schema change I describe, following the safety rule exactly.
List the lock risk and the app code that must change alongside it.

15. Type-check must be clean before PR

---
description: No any, @ts-ignore, or non-null assertion without an inline justification comment
globs: "**/*.ts,**/*.tsx"
alwaysApply: true
---
- `any`, `@ts-ignore`, and `!` assertions each require a `// why:` comment on the same line.
- A suppression without that comment should be treated as an error to fix, not merge.
/fix-types
Run the type checker and fix every error.
Where a suppression is genuinely needed, add the `// why:` comment the rule requires instead of silencing it bare.

16. Changelog entries are mandatory for user-facing changes

---
description: User-facing changes require a CHANGELOG.md entry in the same PR
globs: "app/**/*.tsx,app/**/*.ts"
alwaysApply: false
---
- If the diff changes behavior a user can see, CHANGELOG.md needs a new line under Unreleased.
- Internal refactors and test-only changes are exempt.
/changelog
Update CHANGELOG.md with the changes since the last entry, per the rule (skip internal-only changes).
Use commit history as the source; do not invent entries.

17. API docs stay in sync with route handlers

---
description: Public API route changes require the matching OpenAPI/docs entry to be updated in the same PR
globs: app/api/v1/**/*.ts
alwaysApply: false
---
- A new or changed public endpoint updates the OpenAPI spec file in the same change.
- Breaking changes to a response shape need a version note, not a silent change.
/api-docs-sync
Compare app/api/v1 route handlers against the OpenAPI spec file.
List any endpoint whose spec is out of date and update the spec to match the code.

18. README stays runnable

---
description: README setup steps must match package.json scripts and actually work from a clean clone
globs: "README.md,package.json"
alwaysApply: false
---
- Every command in the README's setup section must exist as a script in package.json.
- Env vars referenced in setup must be listed in .env.example.
/readme-check
Verify the README setup steps against package.json scripts and .env.example per the rule.
Fix any mismatch — a command that doesn't exist, an env var that's undocumented.

19. Dependency additions get a reason

---
description: New dependencies in package.json must be justified — no adding a library for one function
globs: package.json
alwaysApply: false
---
- Before adding a dependency, check if the function it provides is under ~20 lines to hand-write.
- Note the reason for the dependency in the PR description, not just the diff.
/dep-review
Review any new entries in package.json against the rule.
For each, tell me whether it's justified or whether the functionality should be hand-written instead.

20. Error logging must include context

---
description: Caught errors must be logged with request/user context, not just the error object
globs: "app/api/**/*.ts,lib/**/*.ts"
alwaysApply: false
---
- `catch` blocks that log must include at least one identifier (userId, requestId, or route) alongside the error.
- Never log a raw error to the client response; log server-side and return a generic message.
/error-handling-audit
Check catch blocks in the files I point at against the logging rule.
Flag any that log bare errors without context or that leak error detail to the client response.

5 advanced chaining combos

These feed one command's output into the next, with a rule anchoring the chain so later steps don't drift from the stack's conventions.

21. Schema change → migration → types → tests

Rule: the migration-safety rule from #14, scoped to migrations/**. Chain: /db-migration (writes the migration) → /fix-types (the generated types from the ORM now need the codebase to catch up) → /write-tests (cover the new column's validation). Run them in that order in the same session so each command sees the previous one's diff.

22. Security pass → fix → re-audit

Rule: the auth rule from #3. Chain: /security-pass (find gaps) → apply the fixes it suggests → /security-pass again on the same files to confirm the second pass reports clean. Don't skip the re-run; the first pass can introduce a new gap while fixing another (e.g., adding a tenant filter that breaks an admin override path).

23. Webhook audit → idempotency fix → release notes

Rule: the webhook rule from #1. Chain: /webhook-audit → fix the flagged handlers → /release-notes scoped to just that range of commits, so the fix gets documented as a reliability improvement instead of disappearing into a generic "bug fixes" line.

24. Pricing audit → checkout review → PR description

Rule: pricing rule (#5) and CRO rule (#6) both active since they share the app/(marketing) and app/checkout trees. Chain: /pricing-audit → /checkout-review on the same release branch → /pr-description, so the PR description the agent writes already knows both audits passed and can say so instead of you writing it manually.

25. CI triage → fix → changelog

Rule: the CI rule (#13) and the changelog rule (#16). Chain: /ci-triage (classify the red run) → fix the real regression → /changelog, scoped to that fix. This is the pattern for a solo founder shipping fast: one failing build becomes one documented fix instead of a silent force-merge.

How AgenticKit replicates these in one run

Hand-writing 25 rule/command pairs and keeping them in sync as your stack changes is the actual cost of this pattern — not the combos themselves, which are mostly variations on "state the invariant once, reference it from the action." AgenticKit's generators are built around that split: the Cursor Rules Generator and Cursor Commands Builder both read your repo's actual stack before writing anything, so a rule it writes for your billing code references your real file paths and your real ORM, and the matching command can safely assume the rule is loaded rather than re-explaining the constraint.

Running npx agentickit-cursor init --target cursor installs 47 commands and their paired rules together, plus the agents and skills that back the heavier ones (a security-auditor agent behind /security-pass, a billing-engineer agent behind /billing-fix). You don't have to write the pairing logic yourself — it's already wired as agent → skill → command → rule, which is the same four-primitive stack covered in Rules vs Skills vs Commands vs Hooks.

FAQ

Do I need a rule for every command? No. Commands like /pr-description or /standup don't touch a standing invariant, so there's nothing to pair them with. Pair only commands that assume a guardrail — security, billing state, migration safety — where running the command without the rule loaded could produce something wrong.

Will the rule and the command conflict if I edit one and not the other? Only if the command's prompt re-states the rule's logic instead of referencing it. Write the command to assume the rule, not duplicate it, and they can't drift out of sync with each other — though both can still drift from the actual code, which is why the audit-style commands (/webhook-audit, /security-pass) exist to catch that.

Does this pattern work the same way in Claude Code? Yes — swap .cursor/rules/*.mdc for a skill with a clear description (skills load automatically, same as a glob-matched rule) and .cursor/commands/*.md for .claude/commands/*.md. The pairing logic is identical; only the file format and activation mechanism differ. See Cursor Skills vs Claude Code Skills for the activation differences.

Where should I start if I only have time for 5? Combos #1, #3, #11, #14, and #15 — webhook idempotency, server-action auth, skipped-test tracking, migration safety, and type-check cleanliness. Those four guard the mistakes that are hardest to notice in review and most expensive to ship.

Sources

Written by Rakshit Yadav. Published 2026-10-31, last updated 2026-10-31.

Related Tools & Agents

🛠️ Free Tool: cursor-rules-generator🛠️ Free Tool: cursor-commands-builder🛠️ Free Tool: rules-grader🤖 Agent: code-reviewer🤖 Agent: security-auditor🤖 Agent: billing-engineer⚡ Command: /security-pass⚡ Command: /db-migration⚡ Command: /billing-fixSkill: project-conventionsSkill: security-basicsSkill: billing-dunning

Get commands and rules paired already

47 commands land with their matching rules in .cursor/ and .claude/ with one install.

Browse commands →