Summary
Stored agent memory records an incorrect generalization about Cobertura root-attribute deduplication: the claim holds only for raw dotnet-coverage output, not for post-processed ConvertTo-KoverageCoberturaXml artifacts.
Environment
- OS/version: Windows 11 Pro 10.0.26200
- Python version: n/a (documentation defect)
- Command/flags used: n/a
- Data source or fixture:
.claude/agent-memory/atomic-executor/project_coverage_delta_reproduce_baseline_counting_method.md:34-36
Steps to Reproduce
- Read
.claude/agent-memory/atomic-executor/project_coverage_delta_reproduce_baseline_counting_method.md:34-36.
- Apply its stated rule to a post-processed
ConvertTo-KoverageCoberturaXml artifact.
- Compare the resulting figure against the class-level rollup count.
Expected Behavior
The memory should state the distinction between raw generator output and post-processed artifacts explicitly, so a future agent does not skip a needed adjustment.
Actual Behavior
The memory asserts that the repository-wide root <coverage> attributes "are already deduped and match a per-package all-descendant sum in this repo, so repo-level figures need no adjustment." That is true only of raw dotnet-coverage output. It was false for any post-processed ConvertTo-KoverageCoberturaXml artifact, where (before #441 was fixed) the root attributes were the all-descendant sum - the very defect #441 corrects.
Logs / Screenshots
Impact / Severity
A future agent following the memory verbatim could skip a needed adjustment when reading a post-processed artifact. No production code is affected.
Source
From: docs/features/potential/2026-08-10-agent-memory-cobertura-dedup-generalization-wrong.md
Summary
Stored agent memory records an incorrect generalization about Cobertura root-attribute deduplication: the claim holds only for raw
dotnet-coverageoutput, not for post-processedConvertTo-KoverageCoberturaXmlartifacts.Environment
.claude/agent-memory/atomic-executor/project_coverage_delta_reproduce_baseline_counting_method.md:34-36Steps to Reproduce
.claude/agent-memory/atomic-executor/project_coverage_delta_reproduce_baseline_counting_method.md:34-36.ConvertTo-KoverageCoberturaXmlartifact.Expected Behavior
The memory should state the distinction between raw generator output and post-processed artifacts explicitly, so a future agent does not skip a needed adjustment.
Actual Behavior
The memory asserts that the repository-wide root
<coverage>attributes "are already deduped and match a per-package all-descendant sum in this repo, so repo-level figures need no adjustment." That is true only of rawdotnet-coverageoutput. It was false for any post-processedConvertTo-KoverageCoberturaXmlartifact, where (before #441 was fixed) the root attributes were the all-descendant sum - the very defect #441 corrects.Logs / Screenshots
coverage-baseline.cobertura.xml: class-level<line>count 79957 equals its ownlines-valid="79957"; the all-descendant count is 161086.coverage-final.cobertura.xml: the all-descendant count 110849 equals its emittedlines-valid="110849"; the class-level count is 62345.Impact / Severity
A future agent following the memory verbatim could skip a needed adjustment when reading a post-processed artifact. No production code is affected.
Source
From: docs/features/potential/2026-08-10-agent-memory-cobertura-dedup-generalization-wrong.md