Forge
markdowne8ad0934
1# .NET Roslyn and Debug Playbook v1
2
3## Related
4
5- Обязательный чеклист агента Cursor (Roslyn MCP, `file_path` по каждому затронутому `.cs`): [`playbook-csharp-roslyn-mcp-diagnostics-v1.md`](playbook-csharp-roslyn-mcp-diagnostics-v1.md).
6
7## Purpose
8Specialized playbook for C# diagnostics, refactorings, and debug workflows in .NET projects.
9
10## Scope
11- Roslyn-first diagnostics and code actions
12- Structured build/test validation
13- Debug session orchestration and evidence capture
14- Fast root-cause isolation for C# runtime and logic failures
15
16## Evidence-Based Working Format
17- Fact: capture the exact failing diagnostic, test, or runtime symptom.
18- Hypothesis: define the minimal likely cause.
19- Check: apply smallest fix and re-run the shortest validating loop.
20- Decision criterion: predefine close/escalate threshold.
21- Confidence mark: explicit certainty tag and unresolved risk.
22
23## Roslyn-First Contracts
24- Use Roslyn diagnostics as primary source for C# errors/warnings.
25- Prefer `get_code_actions` + `apply_code_action` over manual edits when available.
26- Use symbol-aware navigation (`go_to_definition`, `find_usages`) for refactoring.
27- Use semantic rename (`roslyn_rename`) for identifier-safe renames.
28
29## Diagnostic Loop
30- Scope diagnostic pass: file -> project -> solution.
31- Fix highest-severity issues first, then warnings with behavior risk.
32- Re-run diagnostics after each meaningful batch.
33- Track recurring warning classes into checklist updates.
34
35## Debug Session Contracts
36- Always make target and arguments explicit before launch.
37- Keep breakpoints tied to current target artifact.
38- Use continue/step/inspect cycle with explicit assumption checks.
39- End session with concise root cause + fix verification note.
40
41## Build/Test Validation Contracts
42- Run build after structural refactors.
43- Run impacted tests before broad test suites.
44- Treat flaky tests as defects in feedback reliability.
45- Keep local loop aligned with CI contracts.
46
47## Common Failure Patterns
48- Symbol rename without usage sweep -> semantic drift.
49- Manual fix bypassing code action -> hidden style/semantic inconsistency.
50- Debug without stable repro -> misleading observations.
51- Multi-fix commit without intermediate checks -> unclear causality.
52
53## Metrics
54- Time to first valid diagnosis.
55- Time to green build after first failure.
56- Ratio of code-action fixes vs manual fixes.
57- Reopen rate for fixed debug incidents.
58
59## Revisit Triggers
60- Repeat diagnostic classes over multiple cycles.
61- High debug time with weak reproducibility.
62- Frequent mismatch between local and CI outcomes.
63- Rising regressions after refactors.
64
65
View only · write via MCP/CIDE