il-copy-inspection

Sterne6

Forks1

Aktualisiert10. Juli 2026 um 05:41

Find struct value copies in a method's compiled IL - defensive copies, boxing, by-value field/argument/return copies - by reading the emitted bytecode rather than predicting from source. Use when asked to "find struct copies", "where does the compiler copy this struct", "is this a defensive copy", "check for boxing in IL", "did the compiler emit a copy here", "confirm the analyzer's defensive-copy warning", or "audit a [NonCopyable] type's copies after build". Post-build, ground-truth counterpart to source-level defensive-copy analyzers - IL is post-lowering, so synthesized copies the analyzer cannot see are visible here. Not for wall-clock/allocation measurement (that is `performance-testing`) nor for JIT-emitted machine code (that is `framework-jit-optimization` + `DisassemblyDiagnoser`).

Installation

Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.

Quelle

JeremyKuhne

JeremyKuhne/touki

GitHub-Repository öffnen Creator-Repositorys ansehen

Download

Verwandte BerufeSOC

Basierend auf der SOC-Berufsklassifikation

SoftwareentwicklerInformatik- und Mathematikberufe·SOC 15-1252

Datei-Explorer

3 Dateien

SKILL.md

readonly

Mehr aus diesem Repository

gleiches Repository

filtrace

JeremyKuhne/touki

Analyze .NET CPU, allocation, exception, GC, JIT, and wall-clock (thread-time) data in .nettrace, .etl, and speedscope files with the filtrace CLI or MCP server. Use when a user asks where time or allocation volume goes in a trace or benchmark, which method or source line is hot, why a run regressed against a baseline, what a captured .nettrace / .etl contains, or to rank / drill / diff / export a profile - including profiling .NET Framework (net481) via ETW, where an EventPipe ranking would mislead. Also covers capturing the trace first - choosing EventPipe vs ETW, elevation, and the recording tool (dotnet-trace, BenchmarkDotNet, PerfView, wpr).

2026-07-166

performance-testing

JeremyKuhne/touki

Author and run BenchmarkDotNet performance tests in a multi-targeted .NET library's perf project, and translate a user's outcome-shaped performance question into a measurement. Use when adding new benchmarks, running existing ones, comparing implementations, profiling to find which method or source line dominates, evaluating allocations / memory usage, reading the generated code (sharplab / DisassemblyDiagnoser / HardwareCounters), or when a user asks how long something takes, how much memory it uses, where time is spent, or to help make a method faster - which this skill turns into a scenario, a benchmark, and a drill-down.

2026-07-166

agent-files-review

JeremyKuhne/touki

Review changes to AI-agent customization files (AGENTS.md, .github/copilot-instructions.md, *.instructions.md, *.prompt.md, *.agent.md, SKILL.md, the validator, the CI workflow). Use when asked to review or validate agent-file changes, fix CI failures from the agent-files workflow, or audit a draft of any of these files.

2026-07-156

address-pr-feedback

JeremyKuhne/touki

Address feedback on an existing pull request - review comments, requested changes, CI failures, or any post-PR follow-up work. Use when the user says "address the review", "fix the comments", "address Copilot's feedback", "fix the CI failure", or any similar phrasing. Distinct from `create-pr`, which covers opening the *initial* PR.

2026-07-106

create-pr

JeremyKuhne/touki

Create a pull request for the current changes. Use when asked to "make a PR", "open a pull request", "push and PR", or otherwise publish in-progress work for review. Ensures changes are on a non-`main` branch, commits are made and pushed, and the PR targets `upstream/main` when an `upstream` remote exists, otherwise `origin/main`.

2026-07-106

code-comprehension

JeremyKuhne/touki

Evidence-based readability and cognitive-load review of code. Use when asked to "review this for readability", "is this too complex", "reduce nesting or cognitive load", "what is a reasonable method length / parameter count / nesting depth", or when judging whether code will be hard to understand. Screens code against research-backed thresholds and prioritizes the factors that actually drive comprehension - naming first, then structure that loads working memory.

2026-07-106

compatibility	Requires a Release build with portable or embedded PDBs and an IL reader such as ilspycmd, ildasm, Cecil, or System.Reflection.Metadata.
description	Find struct value copies in a method's compiled IL - defensive copies, boxing, by-value field/argument/return copies - by reading the emitted bytecode rather than predicting from source. Use when asked to "find struct copies", "where does the compiler copy this struct", "is this a defensive copy", "check for boxing in IL", "did the compiler emit a copy here", "confirm the analyzer's defensive-copy warning", or "audit a [NonCopyable] type's copies after build". Post-build, ground-truth counterpart to source-level defensive-copy analyzers - IL is post-lowering, so synthesized copies the analyzer cannot see are visible here. Not for wall-clock/allocation measurement (that is `performance-testing`) nor for JIT-emitted machine code (that is `framework-jit-optimization` + `DisassemblyDiagnoser`).
license	MIT
metadata	{"applicability":"dotnet","binding":"optional-overlay","github-path":"skills/il-copy-inspection","github-pinned":"v0.10.0","github-ref":"refs/tags/v0.10.0","github-repo":"https://github.com/JeremyKuhne/agent-skills","github-tree-sha":"07bce64dc12f798aeeec01aab716d26bbde5c0ed","maturity":"canary","portability":"portable","related":"roslyn-analyzers, framework-jit-optimization, performance-testing, scratch-buffer-strategy","requires":"none","risk":"advisory"}
name	il-copy-inspection

Inspecting IL for struct copies

If overlay.md exists beside this file, read it before acting; it contains repository-specific bindings. This core remains usable without it.

A struct copy is invisible in C# source but concrete in IL. This skill reads a method's emitted bytecode to find where the compiler actually copies a value type - defensive copies, boxing, and by-value field/argument/return copies - and maps each back to a source line. It is the ground-truth, post-build complement to the predictive, in-IDE rules of a source-level defensive-copy analyzer.

The four layers (where this sits)

A struct copy can be reasoned about at four levels. Pick the layer that answers the question; they are complementary, not interchangeable.

Layer	Artifact	Tool / skill	What it tells you
Source	`IOperation` (pre-lowering)	`roslyn-analyzers` (defensive-copy rules)	A copy is predicted from C# rules, live in the IDE
IL	compiled bytecode (post-lowering)	this skill (`ildasm` / `ilspycmd` / Cecil)	*A copy was emitted* by the C# compiler**
Asm	JIT machine code	`framework-jit-optimization` + `DisassemblyDiagnoser`	Whether the JIT kept the copy or elided it
Runtime	ETW / wall-clock	`performance-testing` + `filtrace`	Whether the copy costs measurable time / allocation

The IL layer is uniquely able to see copies the compiler synthesizes during lowering - async / iterator state-machine field hoisting, closures, thunks - which have no source operation and are therefore out of reach for an analyzer. It is also the cheapest way to confirm a defensive-copy prediction: if the IL has the defensive-copy signature, the analyzer was right.

Step 0 - is something already answering this?

"Does my code have a defensive copy I can fix in the editor?" -> a source-level defensive-copy analyzer (roslyn-analyzers) already flags the common cases live. Use this skill only to (a) confirm a flagged case, or (b) find the synthesized copies the analyzer documents as out of scope.
"Is this copy actually costing me time / allocations?" -> performance-testing (boxing shows as Allocated under [MemoryDiagnoser]; a hot defensive copy shows in a trace).
"Did the JIT keep the copy?" -> framework-jit-optimization / [DisassemblyDiagnoser]. The C# compiler may emit an IL copy that the JIT then elides, so IL presence is necessary but not sufficient for a runtime cost.
Otherwise, read the IL (below).

Workflow

Build Release with a portable PDB. Optimized IL is what ships; Debug IL has extra copies and spills. The PDB's sequence points are what map IL offsets back to file:line. Make sure the build emits a portable PDB - <DebugType>embedded</DebugType> or <DebugType>portable</DebugType> in the project.
Extract the method's IL. Pick a tool from the table below. Disassemble the single method of interest, not the whole assembly.
Match the copy patterns. Scan the method body for the copy opcodes and the defensive-copy temp signature. The full catalog - every opcode, the ldobj/stloc/ldloca/call defensive sequence, box, cpobj, ldfld vs ldflda - is in references/copy-opcodes.md.
Map IL offset to source. Use the PDB sequence points (most disassemblers can interleave source, e.g. ilspycmd -il with debug info, or read sequence points via System.Reflection.Metadata) to attribute each copy to a line - the same artifact-to-source drill the performance-testing skill does with filtrace.
Report the type copied, the mechanism (defensive / box / by-value), and the source line. Distinguish intended copies (a documented value transfer) from unintended ones; IL alone cannot, so use the source context.

Tool options

Tool	Good for	Note
`ilspycmd -il <dll>` (ICSharpCode.ILSpyCmd)	One-method IL dump, scriptable	`dotnet tool install -g ilspycmd`; cross-platform
`ildasm`	Quick Windows dump	Ships with the .NET SDK / VS; Windows-centric
`Mono.Cecil`	Programmatic body walking, custom rules	Best when scripting a repeatable copy audit
`System.Reflection.Metadata` (`MetadataReader`, `MethodBodyBlock`)	In-proc, no extra dependency, owns the PDB sequence-point mapping	Most code; the right base for an automated pass

For a repeatable, automated audit, Mono.Cecil or System.Reflection.Metadata walking the method body and matching the opcode patterns from the catalog is the durable choice; ilspycmd is the fastest manual spot-check.

Constraints

Needs a build and a PDB. Unlike an analyzer (source only), this requires compiled output. No PDB means no source-line mapping - opcodes only.
IL copy != runtime cost. The JIT may elide an emitted copy. Confirm cost at the asm or runtime layer before claiming a perf impact.
Generics share IL. A copy of T appears once in the shared body; the real cost depends on the instantiation. Note the open generic, then reason per value-type substitution.
Intent is lost in IL. A by-value copy that is a deliberate transfer looks identical to an accidental one. Keep the source open to classify; this is the same "unintended vs intended" limit a defensive-copy analyzer's docs call out.

Cross-skill

roslyn-analyzers - the source-level, predictive side (defensive-copy rules). This skill confirms its predictions and covers the synthesized copies it cannot see.
framework-jit-optimization - the next layer down (asm); whether the JIT kept the copy, and what to do about it on .NET Framework.
performance-testing - whether the copy costs measurable time / allocation.
scratch-buffer-strategy - many [NonCopyable] types are pooled buffers; this skill audits whether one is being copied by value.
A consuming repository wires these cross-references and its concrete diagnostic IDs and project names in its overlay.

Disambiguation

"Find the copies" is ambiguous across layers. Route by artifact: predict from source, live -> roslyn-analyzers; read the compiler's emitted IL -> this skill; read the JIT's machine code -> framework-jit-optimization; measure the runtime cost -> performance-testing. This skill never runs the code and never measures time - it reads static bytecode.