Expecho/Agent42 / .claude/skills/dotnet-complexity
dotnet-complexity
UnverifiedReport cyclomatic complexity of C#/.NET code — changed files by default. Uses Roslyn code metrics and falls back to lizard. Optional, model-invoked only.
.NET cyclomatic complexity
Read-only report — writes no files into the repo. Ranks methods by cyclomatic
complexity so hotspots surface before review. OPTIONAL: lives in
.agentic/optional-skills/; onboarding copies it into .claude/skills/ only when
the repo is .NET and carries a cyclomatic_complexity_max gate in
.agentic/quality-gates.yaml. This is the measuring tool that /execute-plan
step 4 runs to produce the actual value for that gate. Requires the .NET SDK —
check dotnet --version.
Step 1 — Resolve scope
-
Argument given (path, project, or glob): analyze exactly that. Done.
-
No argument: analyze the changed
.csfiles vs the base branch. Build the set from committed, staged, and unstaged changes; drop generated files (*.g.cs,*.Designer.cs) and anything underbin/,obj/,node_modules/:BASE="${BASE_BRANCH:-main}" # the branch this work merges into MERGE_BASE=$(git merge-base HEAD "$BASE") { git diff --name-only --diff-filter=ACMR "$MERGE_BASE"...HEAD git diff --name-only --diff-filter=ACMR git diff --name-only --diff-filter=ACMR --cached } | sort -u | grep -E '\.cs$' | grep -vE '\.(g|Designer)\.cs$' | grep -vE '(^|/)(bin|obj|node_modules)/'If the set is empty, say so and stop — there is nothing to measure.
Step 2 — Pick the tool (Roslyn metrics, fall back to lizard)
Roslyn metrics run per project/solution (not per file), so measure the owning project(s) and post-filter to the changed files from Step 1.
-
Roslyn code metrics (primary) — the official Microsoft path. Emits every method's cyclomatic complexity regardless of any threshold:
# Run the built-in Metrics MSBuild target against the project/solution that owns # the changed files. It writes an XML report (path from the /p:MetricsOutputFile # value you pass, or the target's default location). dotnet build <project-or-solution> /t:Metrics /p:MetricsOutputFile=<scratchpad>/metrics.xmlParse the XML: each
<Method>carriesCyclomaticComplexity; keep the ones in changed files. If the Metrics target is unavailable (older SDK / not wired up), the project can referenceMicrosoft.CodeAnalysis.Metrics— resolve the version from the feed, never from memory (hard rule 4) — or use the fallback below. Write the XML to the session scratchpad, never into the repo tree. -
lizard (fallback) — a dedicated cross-language CC tool, when the Metrics target isn't available or you want a no-MSBuild read:
pip install --quiet lizard # once; or pipx run lizard lizard <files> # per-method CCN, takes explicit paths
State which path you took and why (one line) so the numbers are interpretable.
Step 3 — Report
- A table ranked worst-first: file · method/line · cyclomatic complexity.
- Threshold comes from config, never memory (hard rule 3). Read
gates.cyclomatic_complexity_maxfrom.agentic/quality-gates.yaml.- If it exists: flag every method at or over it, and state the source.
- If it does not exist: report raw numbers, state plainly that no cyclomatic gate
is configured, and offer to add one to
quality-gates.yaml(do not invent a pass/fail line from a hardcoded number).
In THIS repo that key does not exist, and that is deliberate — do not "fix" it by
scoring against cognitive_complexity_max (ADR 0050). Cyclomatic and cognitive
complexity are different metrics with different numbers: ChatCoreAsync measures 29
cyclomatic and 39 cognitive, and this repo's gate is the cognitive one, measured at
build time by SonarAnalyzer.CSharp rule S3776 — not by this skill. So this skill is
informational here, never the row /execute-plan step 4 renders. Report raw
cyclomatic numbers, say which metric they are, and leave the gate to the build.
- Keep it to the report. Refactoring is a separate, explicit ask.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| receiving-code-reviewobra/superpowersUse when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation | Stars: 318k | Updated | Claude CodeCodex | Unverified | Install |
| requesting-code-reviewobra/superpowersUse when completing tasks, implementing major features, or before merging to verify work meets requirements | Stars: 318k | Updated | Claude CodeCodex | Unverified | Install |
| using-git-worktreesobra/superpowersUse when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallback | Stars: 318k | Updated | Claude CodeCodex | Unverified | Install |
| verification-before-completionobra/superpowersUse when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always | Stars: 318k | Updated | Claude CodeCodex | Unverified | Install |