
prowler-api
Prowler API patterns: RLS, RBAC, providers, Celery tasks. Trigger: When working in api/ on models/serializers/viewsets/filters/tasks involving tenant isolation (RLS), RBAC, or provider lifecycle.
- Installs
- 73
- Last 24h
- —
- Trending rank
- —
- Security
- Not audited
Install
Works with every agent the skills CLI supports.
npx skills add https://github.com/prowler-cloud/prowler --skill prowler-api- 01
Install
Run the command in your project. The CLI asks which agents to install for.
- 02
Describe the task
Ask your agent for something this skill covers. It loads prowler-api automatically when relevant.
- 03
Review the result
The agent follows the SKILL.md steps, scripts and references bundled with the skill.
SKILL.md
The instructions your agent reads when the skill is triggered.
When to Use
Use this skill for Prowler-specific patterns:
- Row-Level Security (RLS) / tenant isolation
- RBAC permissions and role checks
- Provider lifecycle and validation
- Celery tasks with tenant context
- Multi-database architecture (4-database setup)
For generic DRF patterns (ViewSets, Serializers, Filters, JSON:API), use django-drf skill.
Critical Rules
- ALWAYS use
rls_transaction(tenant_id)when querying outside ViewSet context - ALWAYS use
get_role()before checking permissions (returns FIRST role only) - ALWAYS use
@set_tenantthen@handle_provider_deletiondecorator order - ALWAYS use explicit through models for M2M relationships (required for RLS)
- NEVER access
Provider.objectswithout RLS context in Celery tasks - NEVER bypass RLS by using raw SQL or
connection.cursor() - NEVER use Django's default M2M - RLS requires through models with
tenant_id
Note:
rls_transaction()accepts both UUID objects and strings - it converts internally viastr(value).
Architecture Overview
4-Database Architecture
| Database | Alias | Purpose | RLS |
|---|---|---|---|
default | prowler_user | Standard API queries | Yes |
admin | admin | Migrations, auth bypass | No |
replica | prowler_user | Read-only queries | Yes |
admin_replica | admin | Admin read replica | No |
# When to use admin (bypasses RLS)
from api.db_router import MainRouter
User.objects.using(MainRouter.admin_db).get(id=user_id) # Auth lookups
# Standard queries use default (RLS enforced)
Provider.objects.filter(connected=True) # Requires rls_transaction context
RLS Transaction Flow
Request → Authentication → BaseRLSViewSet.initial()
│
├─ Extract tenant_id from JWT
├─ SET api.tenant_id = 'uuid' (PostgreSQL)
└─ All queries now tenant-scoped
Implementation Checklist
When implementing Prowler-specific API features:
| # | Pattern | Reference | Key Points |
|---|---|---|---|
| 1 | RLS Models | api/rls.py | Inherit RowLevelSecurityProtectedModel, add constraint |
| 2 | RLS Transactions | api/db_utils.py | Use rls_transaction(tenant_id) context manager |
| 3 | RBAC Permissions | api/rbac/permissions.py | get_role(), get_providers(), Permissions enum |
| 4 | Provider Validation | api/models.py | validate_<provider>_uid() methods on Provider model |
| 5 | Celery Tasks | tasks/tasks.py, api/decorators.py, config/celery.py | Task definitions, decorators (@set_tenant, @handle_provider_deletion), RLSTask base |
| 6 | RLS Serializers | api/v1/serializers.py | Inherit RLSSerializer to auto-inject tenant_id |
| 7 | Through Models | api/models.py | ALL M2M must use explicit through with tenant_id |
Full file paths: See references/file-locations.md
Decision Trees
Which Base Model?
Tenant-scoped data → RowLevelSecurityProtectedModel
Global/shared data → models.Model + BaseSecurityConstraint (rare)
Partitioned time-series → PostgresPartitionedModel + RowLevelSecurityProtectedModel
Soft-deletable → Add is_deleted + ActiveProviderManager
Which Manager?
Normal queries → Model.objects (excludes deleted)
Include deleted records → Model.all_objects
Celery task context → Must use rls_transaction() first
Which Database?
Standard API queries → default (automatic via ViewSet)
Read-only operations → replica (automatic for GET in BaseRLSViewSet)
Auth/admin operations → MainRouter.admin_db
Cross-tenant lookups → MainRouter.admin_db (use sparingly!)
Celery Task Decorator Order?
@shared_task(base=RLSTask, name="...", queue="...")
@set_tenant # First: sets tenant context
@handle_provider_deletion # Second: handles deleted providers
def my_task(tenant_id, provider_id):
pass
RLS Model Pattern
from api.rls import RowLevelSecurityProtectedModel, RowLevelSecurityConstraint
class MyModel(RowLevelSecurityProtectedModel):
# tenant FK inherited from parent
id = models.UUIDField(primary_key=True, default=uuid4, editable=False)
name = models.CharField(max_length=255)
inserted_at = models.DateTimeField(auto_now_add=True, editable=False)
updated_at = models.DateTimeField(auto_now=True, editable=False)
class Meta(RowLevelSecurityProtectedModel.Meta):
db_table = "my_models"
constraints = [
RowLevelSecurityConstraint(
field="tenant_id",
name="rls_on_%(class)s",
statements=["SELECT", "INSERT", "UPDATE", "DELETE"],
),
]
class JSONAPIMeta:
resource_name = "my-models"
M2M Relationships (MUST use through models)
class Resource(RowLevelSecurityProtectedModel):
tags = models.ManyToManyField(
ResourceTag,
through="ResourceTagMapping", # REQUIRED for RLS
)
class ResourceTagMapping(RowLevelSecurityProtectedModel):
# Through model MUST have tenant_id for RLS
resource = models.ForeignKey(Resource, on_delete=models.CASCADE)
tag = models.ForeignKey(ResourceTag, on_delete=models.CASCADE)
class Meta:
constraints = [
RowLevelSecurityConstraint(
field="tenant_id",
name="rls_on_%(class)s",
statements=["SELECT", "INSERT", "UPDATE", "DELETE"],
),
]
Async Task Response Pattern (202 Accepted)
For long-running operations, return 202 with task reference:
@action(detail=True, methods=["post"], url_name="connection")
def connection(self, request, pk=None):
with transaction.atomic():
task = check_provider_connection_task.delay(
provider_id=pk, tenant_id=self.request.tenant_id
)
prowler_task = Task.objects.get(id=task.id)
serializer = TaskSerializer(prowler_task)
return Response(
data=serializer.data,
status=status.HTTP_202_ACCEPTED,
headers={"Content-Location": reverse("task-detail", kwargs={"pk": prowler_task.id})}
)
Providers (11 Supported)
| Provider | UID Format | Example |
|---|---|---|
| AWS | 12 digits | 123456789012 |
| Azure | UUID v4 | a1b2c3d4-e5f6-... |
| GCP | 6-30 chars, lowercase, letter start | my-gcp-project |
| M365 | Valid domain | contoso.onmicrosoft.com |
| Kubernetes | 2-251 chars | arn:aws:eks:... |
| GitHub | 1-39 chars | my-org |
| IaC | Git URL | https://github.com/user/repo.git |
| Oracle Cloud | OCID format | ocid1.tenancy.oc1.. |
| MongoDB Atlas | 24-char hex | 507f1f77bcf86cd799439011 |
| Alibaba Cloud | 16 digits | 1234567890123456 |
Adding new provider: Add to ProviderChoices enum + create validate_<provider>_uid() staticmethod.
RBAC Permissions
| Permission | Controls |
|---|---|
MANAGE_USERS | User CRUD, role assignments |
MANAGE_ACCOUNT | Tenant settings |
MANAGE_BILLING | Billing/subscription |
MANAGE_PROVIDERS | Provider CRUD |
MANAGE_INTEGRATIONS | Integration config |
MANAGE_SCANS | Scan execution |
UNLIMITED_VISIBILITY | See all providers (bypasses provider_groups) |
RBAC Visibility Pattern
def get_queryset(self):
user_role = get_role(self.request.user)
if user_role.unlimited_visibility:
return Model.objects.filter(tenant_id=self.request.tenant_id)
else:
# Filter by provider_groups assigned to role
return Model.objects.filter(provider__in=get_providers(user_role))
Celery Queues
| Queue | Purpose |
|---|---|
scans | Prowler scan execution |
overview | Dashboard aggregations (severity, attack surface) |
compliance | Compliance report generation |
integrations | External integrations (Jira, S3, Security Hub) |
deletion | Provider/tenant deletion (async) |
backfill | Historical data backfill operations |
scan-reports | Output generation (CSV, JSON, HTML, PDF) |
Task Composition (Canvas)
Use Celery's Canvas primitives for complex workflows:
| Primitive | Use For |
|---|---|
chain() | Sequential execution: A → B → C |
group() | Parallel execution: A, B, C simultaneously |
| Combined | Chain with nested groups for complex workflows |
Note: Use
.si()(signature immutable) to prevent result passing. Use.s()if you need to pass results.
Examples: See assets/celery_patterns.py for chain, group, and combined patterns.
Beat Scheduling (Periodic Tasks)
| Operation | Key Points |
|---|---|
| Create schedule | IntervalSchedule.objects.get_or_create(every=24, period=HOURS) |
| Create periodic task | Use task name (not function), kwargs=json.dumps(...) |
| Delete scheduled task | PeriodicTask.objects.filter(name=...).delete() |
| Avoid race conditions | Use countdown=5 to wait for DB commit |
Examples: See assets/celery_patterns.py for schedule_provider_scan pattern.
Advanced Task Patterns
@set_tenant Behavior
| Mode | tenant_id in kwargs | tenant_id passed to function |
|---|---|---|
@set_tenant (default) | Popped (removed) | NO - function doesn't receive it |
@set_tenant(keep_tenant=True) | Read but kept | YES - function receives it |
Key Patterns
| Pattern | Description |
|---|---|
bind=True | Access self.request.id, self.request.retries |
get_task_logger(__name__) | Proper logging in Celery tasks |
SoftTimeLimitExceeded | Catch to save progress before hard kill |
countdown=30 | Defer execution by N seconds |
eta=datetime(...) | Execute at specific time |
Examples: See assets/celery_patterns.py for all advanced patterns.
Celery Configuration
| Setting | Value | Purpose |
|---|---|---|
BROKER_VISIBILITY_TIMEOUT | 86400 (24h) | Prevent re-queue for long tasks |
CELERY_RESULT_BACKEND | django-db | Store results in PostgreSQL |
CELERY_TASK_TRACK_STARTED | True | Track when tasks start |
soft_time_limit | Task-specific | Raises SoftTimeLimitExceeded |
time_limit | Task-specific | Hard kill (SIGKILL) |
Full config: See assets/celery_patterns.py and actual files at
config/celery.py,config/settings/celery.py.
UUIDv7 for Partitioned Tables
Finding and ResourceFindingMapping use UUIDv7 for time-based partitioning:
from uuid6 import uuid7
from api.uuid_utils import uuid7_start, uuid7_end, datetime_to_uuid7
# Partition-aware filtering
start = uuid7_start(datetime_to_uuid7(date_from))
end = uuid7_end(datetime_to_uuid7(date_to), settings.FINDINGS_TABLE_PARTITION_MONTHS)
queryset.filter(id__gte=start, id__lt=end)
Why UUIDv7? Time-ordered UUIDs enable PostgreSQL to prune partitions during range queries.
Batch Operations with RLS
from api.db_utils import batch_delete, create_objects_in_batches, update_objects_in_batches
# Delete in batches (RLS-aware)
batch_delete(tenant_id, queryset, batch_size=1000)
# Bulk create with RLS
create_objects_in_batches(tenant_id, Finding, objects, batch_size=500)
# Bulk update with RLS
update_objects_in_batches(tenant_id, Finding, objects, fields=["status"], batch_size=500)
Security Patterns
Full examples: See assets/security_patterns.py
Tenant Isolation Summary
| Pattern | Rule |
|---|---|
| RLS in ViewSets | Automatic via BaseRLSViewSet - tenant_id from JWT |
| RLS in Celery | MUST use @set_tenant + rls_transaction(tenant_id) |
| Cross-tenant validation | Defense-in-depth: verify obj.tenant_id == request.tenant_id |
| Never trust user input | Use request.tenant_id from JWT, never request.data.get("tenant_id") |
| Admin DB bypass | Only for cross-tenant admin ops - exposes ALL tenants' data |
Celery Task Security Summary
| Pattern | Rule |
|---|---|
| Named tasks only | NEVER use dynamic task names from user input |
| Validate arguments | Check UUID format before database queries |
| Safe queuing | Use transaction.on_commit() to enqueue AFTER commit |
| Modern retries | Use autoretry_for, retry_backoff, retry_jitter |
| Time limits | Set soft_time_limit and time_limit to prevent hung tasks |
| Idempotency | Use update_or_create or idempotency keys |
Quick Reference
# Safe task queuing - task only enqueued after transaction commits
with transaction.atomic():
provider = Provider.objects.create(**data)
transaction.on_commit(
lambda: verify_provider_connection.delay(
tenant_id=str(request.tenant_id),
provider_id=str(provider.id)
)
)
# Modern retry pattern
@shared_task(
base=RLSTask,
bind=True,
autoretry_for=(ConnectionError, TimeoutError, OperationalError),
retry_backoff=True,
retry_backoff_max=600,
retry_jitter=True,
max_retries=5,
soft_time_limit=300,
time_limit=360,
)
@set_tenant
def sync_provider_data(self, tenant_id, provider_id):
with rls_transaction(tenant_id):
# ... task logic
pass
# Idempotent task - safe to retry
@shared_task(base=RLSTask, acks_late=True)
@set_tenant
def process_finding(tenant_id, finding_uid, data):
with rls_transaction(tenant_id):
Finding.objects.update_or_create(uid=finding_uid, defaults=data)
Production Deployment Checklist
Full settings: See references/production-settings.md
Run before every production deployment:
cd api && uv run python src/backend/manage.py check --deploy
Critical Settings
| Setting | Production Value | Risk if Wrong |
|---|---|---|
DEBUG | False | Exposes stack traces, settings, SQL queries |
SECRET_KEY | Env var, rotated | Session hijacking, CSRF bypass |
ALLOWED_HOSTS | Explicit list | Host header attacks |
SECURE_SSL_REDIRECT | True | Credentials sent over HTTP |
SESSION_COOKIE_SECURE | True | Session cookies over HTTP |
CSRF_COOKIE_SECURE | True | CSRF tokens over HTTP |
SECURE_HSTS_SECONDS | 31536000 (1 year) | Downgrade attacks |
CONN_MAX_AGE | 60 or higher | Connection pool exhaustion |
Commands
# Development
cd api && uv run python src/backend/manage.py runserver
cd api && uv run python src/backend/manage.py shell
# Celery
cd api && uv run celery -A config.celery worker -l info -Q scans,overview
cd api && uv run celery -A config.celery beat -l info
# Testing
cd api && uv run pytest -x --tb=short
# Production checks
cd api && uv run python src/backend/manage.py check --deploy
Resources
Local References
- File Locations: See references/file-locations.md
- Modeling Decisions: See references/modeling-decisions.md
- Configuration: See references/configuration.md
- Production Settings: See references/production-settings.md
- Security Patterns: See assets/security_patterns.py
Related Skills
- Generic DRF Patterns: Use
django-drfskill - API Testing: Use
prowler-test-apiskill
Context7 MCP (Recommended)
Prerequisite: Install Context7 MCP server for up-to-date documentation lookup.
When implementing or debugging Prowler-specific patterns, query these libraries via mcp_context7_query-docs:
| Library | Context7 ID | Use For |
|---|---|---|
| Celery | /websites/celeryq_dev_en_stable | Task patterns, queues, error handling |
| django-celery-beat | /celery/django-celery-beat | Periodic task scheduling |
| Django | /websites/djangoproject_en_5_2 | Models, ORM, constraints, indexes |
Example queries:
mcp_context7_query-docs(libraryId="/websites/celeryq_dev_en_stable", query="shared_task decorator retry patterns")
mcp_context7_query-docs(libraryId="/celery/django-celery-beat", query="periodic task database scheduler")
mcp_context7_query-docs(libraryId="/websites/djangoproject_en_5_2", query="model constraints CheckConstraint UniqueConstraint")
Note: Use
mcp_context7_resolve-library-idfirst if you need to find the correct library ID.
FAQ
What does the prowler-api skill do?
Prowler API patterns: RLS, RBAC, providers, Celery tasks. Trigger: When working in api/ on models/serializers/viewsets/filters/tasks involving tenant isolation (RLS), RBAC, or provider lifecycle.
How do I install prowler-api?
Run npx skills add https://github.com/prowler-cloud/prowler --skill prowler-api in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/prowler-api/ (personal) or .claude/skills/prowler-api/ (project).
Which AI agents can use prowler-api?
Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.
Is prowler-api safe to install?
No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.
Who maintains prowler-api?
It is published by prowler-cloud in prowler-cloud/prowler. It has 73 recorded installs.
More from prowler-cloud/prowler
zod-4
prowler-cloud/prowler
Zod 4 schema validation patterns. Trigger: When creating or updating Zod v4 schemas for validation/parsing (forms, request payloads, adapters), including v3 - v4 migration patterns.
zustand-5
prowler-cloud/prowler
Zustand 5 state management patterns. Trigger: When implementing client-side state with Zustand (stores, selectors, persist middleware, slices).
react-19
prowler-cloud/prowler
React 19 patterns with React Compiler. Trigger: When writing React 19 components/hooks in .tsx (React Compiler rules, hook patterns, refs as props). If using Next.js App Router/Server Actions, also use nextjs-16.
tailwind-4
prowler-cloud/prowler
Tailwind CSS 4 patterns and best practices. Trigger: When styling with Tailwind (className, variants, cn()), especially when dynamic styling or CSS variables are involved (no var() in className).
django-drf
prowler-cloud/prowler
Django REST Framework patterns. Trigger: When implementing generic DRF APIs (ViewSets, serializers, routers, permissions, filtersets). For Prowler API specifics (RLS/RBAC/Providers), also use prowler-api.
typescript
prowler-cloud/prowler
Open the skill to read its SKILL.md instructions.
Popular backend & apis skills
find-skills
vercel-labs/skills
Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.
entra-app-registration
microsoft/azure-skills
Guides Microsoft Entra ID app registration, OAuth 2.0 authentication, and MSAL integration. USE FOR: create app registration, register Azure AD app, configure OAuth, set up authentication, add API permissions, generate service principal, MSAL example, console app auth, Entra ID setup, Azure AD authentication. DO NOT USE FOR: Key Vault secrets (use azure-keyvault-expiration-audit), general Azure resource security guidance.
google-agents-cli-adk-code
google/agents-cli
This skill should be used when the user wants to "write agent code", "build an agent with ADK", "add a tool", "create a callback", "define an agent", "use state management" — in a project that needs ADK (Agent Development Kit) API patterns and code examples. It provides a quick reference for agent types, tool definitions, orchestration patterns, callbacks, state management, the graph Workflow API, and reference recipes to study. Do NOT use for scaffolding (use google-agents-cli-scaffold) or deployment (use google-agents-cli-deploy).
embedded-captions
heygen-com/hyperframes
Add captions or subtitles to an existing single-subject talking-head video without editing the footage. Use for plain verbatim captions, cinematic captions embedded behind the subject, VFX captions, “炸/特效/酷炫字幕,” or a named identity from the 35-style catalog. Route by visual identity, not by backend engine. The quiet anchor rail is the default; embed every word only when the user explicitly wants a fully cinematic treatment. The workflow runs locally end to end, including transcription and subject matting; split multi-shot footage before applying it.
entra-agent-id
microsoft/azure-skills
Provision Microsoft Entra Agent Identity Blueprints, BlueprintPrincipals, and per-instance Agent Identities via Microsoft Graph, and configure OAuth 2.0 token exchange (fmipath, OBO, cross-tenant) including the Microsoft Entra SDK for AgentID sidecar. USE FOR: Agent Identity Blueprint, BlueprintPrincipal, agent OAuth, fmipath token exchange, agent OBO, Workload Identity Federation for agents, polyglot agent auth, Microsoft.Identity.Web.AgentIdentities. DO NOT USE FOR: standard Entra app registration (use entra-app-registration), Microsoft Foundry agent authoring (use microsoft-foundry).
neon
neondatabase/agent-skills
Overview of Neon, a complete set of cloud backend primitives around Lakebase Postgres: Auth, Object Storage, Functions, the AI Gateway, and Realtime. Start here to choose Neon for undecided login, files, APIs, and LLM calls, set up the CLI or MCP server, and follow the branch-first workflow. Use when building an app or backend on Neon, or when "Neon" or "Lakebase Postgres" is mentioned. Child skill neon-postgres wins for an existing DATABASEURL, SQL, schema, inspect, or search. Child skill neon-auth wins for login, users, sessions, identity routing, and Managed Better Auth setup. Child skill neon-realtime wins for live queries in reactive apps and agents. Also use for object storage, S3, buckets, serverless functions, function triggers, cron, AI gateway, LLM calls, realtime, logs, Loki, Grafana, observability, postgres, database, backend, Claimable Neon, neon.new, or a no-signup database.