Skip to content

Marketplace

The Komand marketplace is a curated catalog of skills that agents can use. Every skill is reviewed for security and quality before it receives a verified badge — this is the AppExchange model, not the npm wild west.

Komand’s marketplace exists because open-source agent ecosystems have proven that unreviewed plugins are a security liability. Cisco found skills doing data exfiltration, prompt injection vulnerabilities are rampant, and IT departments fight “shadow AI” from unmanaged installs.

The marketplace solves this with:

  • Code review — every skill is reviewed before verification
  • Permission declarations — skills declare what they need upfront
  • User consent — users grant specific permissions when installing
  • Sandboxed execution — skills cannot access resources beyond their declared scope
  • Signed packages — verified skills have a cryptographic signature for integrity

The web dashboard includes a marketplace browser at /marketplace with:

  • Category browsing and search
  • Skill detail pages with description, publisher, version history
  • Ratings and reviews from other users
  • Permission summary before installation
  • One-click install for agents
Terminal window
# Browse all verified skills
curl "https://your-instance/api/skills/?verifiedOnly=true&page=1&pageSize=20"
# Search by category
curl "https://your-instance/api/skills/?category=crm&page=1&pageSize=20"
# Search by publisher
curl "https://your-instance/api/skills/?publisherId=komand-official"
Terminal window
komand skill search "crm contact"
komand skill info crm-contact-lookup
komand skill install crm-contact-lookup --agent default
CategoryExamples
CRMContact lookup, deal management, pipeline analytics
CommunicationEmail send, SMS, push notifications
SchedulingCalendar booking, reminders, availability check
FinanceInvoice generation, payment processing, expense tracking
ResearchWeb search, market data, competitor analysis
ProductivityDocument generation, file management, task tracking
DevelopmentGitHub integration, CI/CD triggers, code review
IndustrialPLC data access, alarm states, site operations

When a user installs a skill for an agent:

  1. The marketplace displays the skill’s required permissions
  2. The user grants or denies each permission
  3. If all required permissions are granted, the skill is added to the agent’s enabled skills
  4. Credentials (API keys, tokens) are configured and stored in the secure vault
Skill requires: ["crm:read", "network:outbound"]
User grants: ["crm:read", "network:outbound"]
Result: ✅ Installed — skill is available to the agent
Skill requires: ["crm:read", "crm:write"]
User grants: ["crm:read"]
Result: ❌ Missing crm:write — installation blocked

Create a publisher account to get your publisher-id and API key:

Terminal window
komand publisher register --name "Your Company" --email [email protected]

Follow the Skill Development guide to implement and test your skill.

Terminal window
dotnet komand skill pack

This bundles your skill into a .komandskill package.

Terminal window
dotnet komand skill publish --api-key YOUR_PUBLISHER_KEY

Your skill enters the review queue with a status of Pending Review.

The Komand team reviews every submission against these criteria:

CriteriaWhat We Check
SecurityNo data exfiltration, proper sandboxing, input validation, no hardcoded secrets
PermissionsMinimal permission requests with justified access for each
QualityClear documentation, comprehensive error handling, test coverage
CompatibilityWorks across supported Komand versions
ContentAccurate description, appropriate naming, no policy violations

Common rejection reasons:

  • Requesting network:outbound without justification
  • Missing error handling for edge cases
  • Insufficient test coverage
  • Vague or misleading skill description

Approved skills receive:

  • Verified badge — displayed in the marketplace catalog
  • Cryptographic signature — SHA-256 hash for package integrity verification
  • Marketplace listing — discoverable by all Komand users
  • Version tracking — updates go through a streamlined re-review

To publish a new version:

  1. Increment the Version in your SkillDefinition
  2. Package and publish as normal
  3. Updates to verified skills go through an expedited review (unless permissions change)
  4. Users on the previous version are notified of the update

Breaking changes to input/output schemas require a major version bump and full re-review.

Skill publishers can charge for their skills and earn revenue:

TierPublisher ShareKomand Share
Standard70%30%
Featured80%20%

Featured status is granted to high-quality, high-demand skills that meet additional criteria (consistent updates, excellent reviews, publisher support).

Free skills are welcome and encouraged — they help grow the ecosystem and establish publisher reputation.

  • Skills that exfiltrate user data beyond their declared scope
  • Skills with undisclosed network access
  • Malware, spyware, or skills with deceptive functionality
  • Skills that violate third-party terms of service
  • Respond to security reports within 48 hours
  • Maintain compatibility with current Komand releases
  • Provide accurate, up-to-date documentation
  • Handle user support for premium skills

Skills can be removed from the marketplace if they:

  • Are found to violate security policies post-verification
  • Become incompatible and the publisher doesn’t update
  • Receive consistent negative reviews without improvement
  • The publisher account is suspended

Komand publishes a set of first-party skills under the komand-official publisher. These serve as reference implementations and cover common use cases:

  • komand-crm — built-in CRM contact and deal operations
  • komand-calendar — scheduling and availability
  • komand-email — email composition and sending
  • komand-web-search — web search and summarisation
  • komand-file-manager — file upload, download, and processing

Official skills are always free and verified. Their source code is included in the SDK as examples.