15 AI Agent Plugin Recipes to Seed Your Marketplace
Ready-to-build plugin specs for every persona in your organization.
Audience: Plugin authors, platform engineers
How to Use This Guide
Each recipe includes:
What it does: the problem it solves
Who it's for: the target persona
Skills needed: what to build
Platforms: where it should work
Difficulty: how hard it is to build
Impact: expected value to the org
Pick 3-5 that match your org's biggest pain points and build those first.
For Software Engineers
1. Code Review Agent
Difficulty: Medium | Impact: Very High
Automates the mechanical parts of code review so humans can focus on design and logic.
| Attribute | Details |
|---|---|
| Skills | check-security, check-standards, check-performance, check-accessibility |
| Platforms | Claude Code, Copilot, Cursor |
| MCP dependencies | None (filesystem only) |
| Key value | Catches 70% of review comments automatically, freeing reviewers for design-level feedback |
Why it works: Code review is the highest-leverage activity for quality, and also the most tedious. Automating the checklist frees humans for the hard parts.
2. Test Writer
Difficulty: Medium | Impact: High
Generates tests that match your project's existing patterns, not generic boilerplate.
| Attribute | Details |
|---|---|
| Skills | generate-unit-tests, generate-integration-tests, generate-regression-tests |
| Platforms | Claude Code, Copilot, Cursor |
| MCP dependencies | None |
| Key value | 3x faster test creation with consistent naming and patterns |
3. Migration Assistant
Difficulty: High | Impact: Very High
Helps engineers migrate between framework versions, languages, or architectures.
| Attribute | Details |
|---|---|
| Skills | analyze-migration-scope, generate-migration-plan, transform-code, validate-migration |
| Platforms | Claude Code, Cursor |
| MCP dependencies | None |
| Example use | Python 2 to 3, React class to hooks, monolith to microservice extraction |
| Key value | Turns 6-month migrations into 6-week projects |
Recipe tip: Seed with your org's most common migration. If everyone is moving from Framework X to Framework Y, build that transformation first.
4. API Design Reviewer
Difficulty: Low | Impact: Medium
Reviews API designs (OpenAPI specs, GraphQL schemas) against org conventions.
| Attribute | Details |
|---|---|
| Skills | check-naming-conventions, check-pagination, check-error-responses, check-versioning |
| Platforms | Claude Code, Copilot, Cursor |
| MCP dependencies | None |
| Key value | Consistent APIs across the organization without a review committee bottleneck |
5. Dependency Auditor
Difficulty: Medium | Impact: High
Audits project dependencies for security vulnerabilities, license issues, and maintenance status.
| Attribute | Details |
|---|---|
| Skills | scan-vulnerabilities, check-licenses, assess-maintenance, suggest-alternatives |
| Platforms | Claude Code, Copilot, Cursor |
| MCP dependencies | Optional: security context MCP for CVE data |
| Key value | Prevents "we shipped with a critical CVE" incidents |
For SRE / On-Call Engineers
6. Incident Responder
Difficulty: High | Impact: Very High
Already included in the boilerplate. Triages incidents, finds TSGs, queries telemetry.
7. Postmortem Writer
Difficulty: Low | Impact: Medium
Generates structured postmortem documents from incident data.
| Attribute | Details |
|---|---|
| Skills | gather-timeline, identify-contributing-factors, generate-action-items, draft-postmortem |
| Platforms | Claude Code, Copilot Studio |
| MCP dependencies | ICM/PagerDuty, Kusto |
| Template | Your org's postmortem template (5 Whys, blameless format, etc.) |
| Key value | Postmortems written in 30 minutes instead of 3 hours |
8. Runbook Automator
Difficulty: High | Impact: Very High
Converts manual runbook steps into semi-automated workflows with human checkpoints.
| Attribute | Details |
|---|---|
| Skills | parse-runbook, generate-automation-steps, insert-human-checkpoints, validate-safety |
| Platforms | Claude Code, Azure AI Foundry |
| MCP dependencies | Infrastructure APIs, deployment systems |
| Key value | Reduces MTTR by automating the "follow these 47 steps" runbooks |
Recipe tip: Start with the 3 most-frequently-used runbooks. High-frequency = high ROI.
For Product Managers / Non-Engineers
9. Sprint Status Agent
Difficulty: Medium | Impact: Medium
Natural language interface for sprint progress, blockers, and team velocity.
| Attribute | Details |
|---|---|
| Skills | summarize-sprint, identify-blockers, compare-velocity, forecast-completion |
| Platforms | Copilot Studio (Teams) |
| MCP dependencies | ADO/Jira |
| Key value | PMs stop asking engineers "what's the status?" and ask the bot instead |
10. Meeting Summarizer
Difficulty: Low | Impact: Medium
Summarizes meeting transcripts into decisions, action items, and owners.
| Attribute | Details |
|---|---|
| Skills | extract-decisions, extract-action-items, identify-owners, generate-summary |
| Platforms | Copilot Studio (Teams) |
| MCP dependencies | Teams/calendar |
| Key value | No more "can you send the notes?" |
For Security / Compliance Teams
11. Threat Model Assistant
Difficulty: High | Impact: Very High
Generates STRIDE threat models from architecture descriptions or code.
| Attribute | Details |
|---|---|
| Skills | identify-trust-boundaries, enumerate-threats, assess-risk, suggest-mitigations |
| Platforms | Claude Code, Copilot |
| MCP dependencies | None |
| Key value | Threat modeling happens at design time, not after the breach |
12. Compliance Checker
Difficulty: Medium | Impact: High
Checks code and infrastructure configs against regulatory requirements.
| Attribute | Details |
|---|---|
| Skills | check-gdpr-compliance, check-data-residency, check-pii-handling, check-audit-logging |
| Platforms | Claude Code, Copilot, Cursor |
| MCP dependencies | None |
| Key value | Compliance shifts left, caught at dev time, not audit time |
Recipe tip: Customize per regulation. A healthcare org needs HIPAA checks. A financial services org needs SOX checks. Don't try to boil the ocean.
For Data / Analytics Teams
13. Query Optimizer
Difficulty: Medium | Impact: Medium
Analyzes SQL/Kusto/Spark queries and suggests performance improvements.
| Attribute | Details |
|---|---|
| Skills | analyze-query-plan, suggest-indexes, rewrite-for-performance, estimate-cost |
| Platforms | Claude Code, Copilot, Cursor |
| MCP dependencies | Optional: Kusto MCP for live query plans |
| Key value | Prevents "that query cost $5,000 in compute" moments |
For New Hires
14. Codebase Navigator
Difficulty: Low | Impact: High
Helps new engineers understand an unfamiliar codebase by answering questions in context.
| Attribute | Details |
|---|---|
| Skills | explain-architecture, trace-data-flow, find-related-code, explain-decision |
| Platforms | Claude Code, Copilot, Cursor |
| MCP dependencies | None (but EngHub/wiki MCP helps for design doc context) |
| Key value | New hires productive in days instead of weeks |
Recipe tip: Seed the instructions with your actual architecture docs. The more context you give, the better the answers.
15. Standards Coach
Difficulty: Low | Impact: Medium
Teaches engineers your org's specific patterns, conventions, and "why we do it this way."
| Attribute | Details |
|---|---|
| Skills | explain-convention, show-pattern-example, suggest-refactor-to-standard |
| Platforms | Claude Code, Copilot, Cursor |
| MCP dependencies | None |
| Key value | The institutional knowledge that usually takes 6 months to absorb |
Prioritization Matrix
| Plugin | Build Difficulty | Impact | Recommended Phase |
|---|---|---|---|
| Code Review Agent | Medium | Very High | Phase 1 (Week 1) |
| Test Writer | Medium | High | Phase 1 (Week 1) |
| Codebase Navigator | Low | High | Phase 1 (Week 1) |
| Incident Responder | High | Very High | Phase 1 (Week 1) |
| Standards Coach | Low | Medium | Phase 2 (Week 5) |
| Postmortem Writer | Low | Medium | Phase 2 (Week 6) |
| Dependency Auditor | Medium | High | Phase 2 (Week 6) |
| Sprint Status Agent | Medium | Medium | Phase 2 (Week 7) |
| API Design Reviewer | Low | Medium | Phase 2 (Week 8) |
| Compliance Checker | Medium | High | Phase 3 (Week 9) |
| Meeting Summarizer | Low | Medium | Phase 3 (Week 10) |
| Threat Model Assistant | High | Very High | Phase 3 (Week 10) |
| Migration Assistant | High | Very High | Phase 3 (Week 11) |
| Query Optimizer | Medium | Medium | Phase 3 (Week 12) |
| Runbook Automator | High | Very High | Post-Phase 3 |
This is a bonus post in the AI Agent Marketplace series.
About the author
- Pushp Vashisht is working as a Software Engineer II at Microsoft, Ireland. For more information pay a visit at: pushp.ovh
Comments
Join the discussionEnterprise AI Agent Marketplace
Most organizations let every team pick their own AI tool. Some use Claude Code, others use GitHub Copilot, others Cursor or Copilot Studio. The result: duplicated workflows, inconsistent governance, and AI capabilities trapped inside individual teams. This series shows you how to fix that with an internal AI Agent Marketplace: a shared catalog of skills, agents, and rules that every team can install into whichever AI platform they already use. Consistency without forced standardization. Inside you'll find a 3-part core walkthrough (why the pattern matters, how to build one, how to integrate it), 5 platform-specific guides (Claude Code, GitHub Copilot, Copilot Studio, Cursor, Azure AI Foundry), and 5 bonus posts covering ROI modeling, a 90-day adoption playbook, 15 ready-to-build plugin recipes, real case studies across company sizes, and an AI maturity model. A production-ready boilerplate repository ships alongside the series so you can fork and customize on day one. Who it's for: platform engineers, engineering managers, and governance teams who want AI adoption to scale without becoming a sprawl of disconnected experiments.
