- Home
- Skills
- Code Quality & Review
- .NET License Exit Planner
More screenshots
Works with the AI Agents you already use
.NET License Exit Planner
Several widely used .NET packages moved from open-source to commercial licensing at a specific major version.
Free
.NET License Exit Planner
Example session with this skill installed
Check LicenseExitLab.slnx for commercial-license NuGet packages with the dotnet-license-exit skill. We want to exit all three and keep the existing behavior exactly. I approve the baseline run, the characterization tests, each migration and each re-run of the gate.
- Read your context and instructions
- Compiled the net license exit
License exit report: LicenseExitLab.slnx
Date: 2026-10-28
SDK used: 10.0.403
Projects: 2 surveyed
Exposure
| Package | Resolved version | Constraint (file:line) | Status | Parent (if transitive) |
|---|---|---|---|---|
| AutoMapper | 15.1.3 | src/LicenseExitLab.App/LicenseExitLab.App.csproj:10 | Exposed now | — (also reaches LicenseExitLab.Tests via ProjectReference) |
| MediatR | 13.1.0 | src/LicenseExitLab.App/LicenseExitLab.App.csproj:11 | Exposed now | — (same) |
| FluentAssertions | 7.2.0 | tests/LicenseExitLab.Tests/LicenseExitLab.Tests.csproj:11 | Upgrade trap | — |
Dependency-bot rules
.github/dependabot.yml:3: NuGet updates with no ignore rule, so the next major bump crosses FluentAssertions 8.0.0.
Configured license keys
None found.
Decisions (user's answers, verbatim)
All three: "exit by migration; preserve existing behavior exactly."
Baseline (before)
dotnet test LicenseExitLab.slnx: exit 0; 5 passed, 0 failed, 0 skipped. TRX: .gate/before.trx.
Characterization tests added
7 tests in tests/LicenseExitLab.Tests/Characterization/: golden JSON (populated and all-null orders), behavior order, notification order, the one-handler-throws case, and two scope-isolation tests. 12/12 green on the old libraries.
Gate result
- FluentAssertions → AwesomeAssertions 9.6.0: passed, 0 differences.
- AutoMapper → Mapperly 4.3.1: two differences found and fixed in the mapper, not the golden file:
TotalthrewArgumentNullExceptionfor a nullLines; AutoMapper returned 0.Linescame backnullinstead of[].- Then passed with 0 differences.
- MediatR → Mediator 3.0.2: the singleton default leaked an order across scopes (5 failures). With
ValidateOnBuild, the container named 10 captive registrations. WithServiceLifetime.Transient, it passed with 0 differences.
Assertion changes in rewritten tests
HaveCountGreaterOrEqualTo→HaveCountGreaterThanOrEqualTo(rename).() => sender.Send(...)→async () => await sender.Send(...)(ValueTask); same subject and expected exception.- No assertion was loosened.
Regressions
None.
Not checked
Interactive approval prompts; Shouldly and plain-services targets; ProjectTo; MediatR stream requests and exception handlers; real-world repos.
Connects securely to your tools. The creator never sees your data.
About this skill
Several widely used .NET packages moved from open-source to commercial licensing at a specific major version. Most teams find out when a dependency bot opens the upgrade PR. This skill finds every exposed package in a solution and states the exposure with evidence: the file:line of the reference, the resolved version, and whether a bot rule will cross the threshold. For the three packages with mature free replacements, it then performs the migration and proves it preserved behavior.
License terms are never hardcoded. Every run fetches the maintainer's own announcement or README and the NuGet license page for the resolved version, and cites the URL, the date and a short quote. The skill quotes the free-tier terms; it never decides whether your organization qualifies. That question is asked and your answer is recorded verbatim.
Use when you want to know whether a .NET solution depends on packages that went commercial, what a license change costs you, or to migrate off AutoMapper, MediatR or FluentAssertions without changing behavior.
Do not use for legal advice on license eligibility, general NuGet updates, vulnerability-only bumps, target-framework upgrades (use the .NET Upgrade Planner), or non-.NET projects.
Why a green build is not enough
The recorded fixture run shows why the gate exists. After the literal AutoMapper → Mapperly port, the code compiled and all 5 of the fixture's own tests passed, yet output had changed. After the MediatR → Mediator port, one of the fixture's tests failed, but only the characterization tests showed what had actually changed:
- A mapping that AutoMapper silently null-guarded threw
ArgumentNullExceptionunder Mapperly. - A null collection that AutoMapper turned into
[]came back asnull. - Mediator's singleton default captured scoped services, so an order stored in one scope was returned from another (the
DbContextcaptive-dependency bug).
The characterization tests the skill writes first are what pinpointed each one. Each difference stopped the migration for a decision instead of being patched away.
How it works
- Survey, read-only, no dotnet, no network. Every
PackageReference, centralPackageVersion,packages.configentry, resolved transitive versions whenproject.assets.jsonexists, and dependency-bot configs (Dependabot, Renovate). - Match against the watchlist, then verify live. Fetch the primary source and NuGet license page for each hit; confirm the last open-source version from NuGet's own version list.
- Classify: Exposed now, Upgrade trap (a plain version or bot rule lets an update cross the threshold), Pinned safely, or Transitive only (names the parent package).
- Ask the license question once per package: license, pin, migrate, or decide later.
- Baseline, on approval. Full build and test with per-test TRX results.
- Plan the exit, ask once. A version pin plus a bot ignore rule, or a migration recipe. AutoMapper and MediatR migrations propose characterization tests as the first write: golden JSON for populated and all-null inputs, behavior and handler order, notification order and the one-handler-throws case, and scope isolation.
- Write on approval, one package at a time. Characterization tests go green on the old library first; then the swap.
- Gate. Same tests by name, same outcomes, no new skips, golden files unedited. A difference is reported as a finding, never fixed by changing the test.
- Report: every license source quoted, every exposure, your decisions verbatim, every write, every run, and everything not checked.
Migration recipes included
- FluentAssertions → AwesomeAssertions (staged namespace change, the 8.0 member renames) or Shouldly.
- AutoMapper → Mapperly or hand-written mapping, with a table of the default behaviors that change results.
- MediatR → Mediator (source-generated) or plain services, covering
ValueTask, pipeline delegates, lifetimes and publishers.
MassTransit, ImageSharp and QuestPDF are detected and reported with pin-or-license options only.
Rules it keeps
- License claims are fetched live and cited; anything that could not be re-verified goes under "Could not verify".
- Not legal advice: free-tier eligibility is your call, and the skill never infers revenue or status.
- Never weakens an assertion, deletes a test, adds a skip or edits a golden file to pass the gate.
- A configured license key is reported by location only, never by value.
Tested
One real run on a synthetic two-project solution (LicenseExitLab: an app library and an xUnit test project) on .NET SDK 10.0.403, net10.0, with AutoMapper 15.1.3, MediatR 13.1.0, FluentAssertions 7.2.0 and a Dependabot config with no ignore rule.
- The survey found all three packages with
file:line, the Dependabot rule, and the AutoMapper and MediatR exposure flowing into the test project through its project reference. - License terms were fetched live from 8 primary pages on 2026-10-28. Last open-source versions were confirmed from NuGet: AutoMapper 14.0.0 (MIT), MediatR 12.5.0 (Apache-2.0), FluentAssertions 7.2.2 (Apache-2.0).
- Baseline 5/5 passed. 7 characterization tests were added; all 12 tests passed on the old libraries.
- FluentAssertions → AwesomeAssertions 9.6.0: 12/12, 0 differences.
- AutoMapper → Mapperly 4.3.1: passed after the gate caught two behavior changes.
- MediatR → Mediator 3.0.2: passed after the gate caught the captive scoped dependency.
- All 5 original tests still pass. Golden files were hash-checked after every step and never edited.
The run also exposed gaps in the skill's own recipes, fixed before packaging, and one weak characterization test, which is recorded honestly. Full details are in docs/VALIDATION.md.
Not tested: interactive approvals (the license decisions were supplied up front), Shouldly and plain-services targets, ProjectTo/EF Core projections, MediatR stream requests and exception handlers, central package management, packages.config, Renovate, MassTransit/ImageSharp/QuestPDF detection, and real-world repos.
What is in the package
Markdown only, no scripts:
SKILL.md.- Eight reference files: survey and inventory, license watchlist, exit options, equivalence gate, three migration recipes, report template.
docs/VALIDATION.md, with the recorded fixture run.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Code Quality & Review.
- Tuned and tested on one synthetic solution written for this purpose.
- The watchlist covers six package families; a package that changed license and is not on it will not be flagged.
- Live fetches depend on the maintainers' pages being reachable; a failed fetch is reported, not filled in from memory.
- Migration recipes cover the common API surface. Uncommon features (stream requests, ProjectTo with runtime parameters) are listed as needing manual checks.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 20 days ago
- Free with an Agensi account
- Software
- .NET SDK
Permissions
Read files: surveys project files, package references and dependency-bot configs. Write files: only version pins, bot ignore rules, characterization tests and migration changes you approve; each write needs a yes. Terminal: dotnet restore, build, test and list package, each asked first. Network: fetches public license sources (project sites, GitHub READMEs, nuget.org license pages); only URLs are requested and no repository content is sent. dotnet restore pulls from your configured NuGet feeds. The package contains Markdown only: no scripts, no binaries.