knowledge: improve review precision from BCApps PR 10252 feedback

[Wenjie Fan (gggdttt)]

Summary

Improves BCQuality knowledge based on maintainer thumbs-down feedback from explicit BCApps PR review runs.

Source feedback

• Adjust audit log tables BCApps#10252 (comment)

Validation

• Frontmatter and article structure validation completed.

Generated by the BC-ALAgentsInternal self-improvement workflow.

Offline evaluation: inconclusive

No knowledge change or no baseline false-positive reproduction. This pair cannot establish target suppression.

Preparation attempt 1: invalid_selection. Coverage requires a patched file and valid lineStart/lineEnd for synthetic__upgrade-alwayslogtable-removal-01.

Preparation attempt 2: validated.

Selection: new. New: synthetic__upgrade-alwayslogtable-removal-01. Reused (complete payloads): ``.

• synthetic__upgrade-alwayslogtable-removal-01 / false_positive_guard / Adjust audit log tables BCApps#10252 (comment): The review bot flagged removal of four table IDs from the OnAfterIsAlwaysLoggedTable allow-list as needing an upgrade tag, and the maintainer rejected this with a THUMBS_DOWN, replying that these are dynamic checks and do not require upgrade. Confirmed against Change Log Management's IsAlwaysLoggedTable: it recomputes the always-logged decision on every insert/modify/delete/rename call and never persists an AlwaysLogTable flag or setup record, so trimming this hard-coded in-code list changes only future auditing decisions, not stored per-company data or schema, and needs no migration or upgrade tag. expected_comments is intentionally empty so no upgrade-migration finding is required in this table-ID range; this is a false-positive guard by omission, not a proof that every possible finding on this hunk is wrong, and it does not assert any severity or recommendation repair beyond the specific upgrade-tag claim that was rejected.
  Coverage references and outcome shapes are validated mechanically. Semantic equivalence, severity calibration and recommendation quality are NOT proved by these checks or a matching F1.

Common engine code: ecf8e31759d6ddd6d78e3a0b7836b40134368009; dataset SHA256: 40F87789093B7D383AE8F4406D8057AC4F15DFD7E86D72D1D3AE9FF599F1692E.
Model: gpt-5.6-luna; judge: gpt-5.3-codex. Exact D: synthetic__upgrade-alwayslogtable-removal-01.
Dataset base: 26365d6a41a96fac41df9dcc8339cb71241fcb1f; candidate: 66098eaf621dcf9a6f3508a2cbe0f7dbe5e06f19.

Arm	Evaluation commit	Engine pin	Knowledge pin
baseline	fa74a78af16283d5bc40f8ff94a703a4cd08e8af	4407f20b44fe92e3693d0444d25cc3430f1d5262	8584217c7506eea7eef27a9db353d78c0cd6a9a1
candidate	57f097a467b133191a2babb7c6b09f061ad701ab	ea0104e6ad2c21173b774ba627852f8c0b65b515	e7ee9b4b5af18f344f5ca45eebeaf92ac578dd69

Baseline

Run: https://github.com/microsoft/BC-Bench/actions/runs/34341668876; conclusion: success; wall clock: 6.4 minutes.
Raw findings and runtime pin proofs are retained in the self-improvement-feedback artifact and the linked evaluation run.

Entry	Expected	Generated	Missed	Unexpected	F1	AI credits	Agent seconds
synthetic__upgrade-alwayslogtable-removal-01	0	0	0	0	1	unavailable	163.777077015

Observed AI credits: unavailable; coverage 0/1 entries. Missing entries are not extrapolated into a total.

Aggregate metric	Value
total	1
expected_comment_count	0
generated_comment_count	0
matched_comment_count	0
missed_comment_count	0
incorrect_comment_count	0
precision	1
recall	1
f1	1

Candidate

Run: https://github.com/microsoft/BC-Bench/actions/runs/34342284452; conclusion: success; wall clock: 7.4 minutes.
Raw findings and runtime pin proofs are retained in the self-improvement-feedback artifact and the linked evaluation run.

Entry	Expected	Generated	Missed	Unexpected	F1	AI credits	Agent seconds
synthetic__upgrade-alwayslogtable-removal-01	0	0	0	0	1	unavailable	195.732926681

Observed AI credits: unavailable; coverage 0/1 entries. Missing entries are not extrapolated into a total.

Aggregate metric	Value
total	1
expected_comment_count	0
generated_comment_count	0
matched_comment_count	0
missed_comment_count	0
incorrect_comment_count	0
precision	1
recall	1
f1	1

Missing telemetry is unavailable, not zero. Evaluation-only metrics exclude candidate generation and are not the full-cycle cost.
Human review must verify source-patch fidelity, gold correctness, and target attribution. No automatic merge or branch-protection claim is made.

[Jesper Schulz-Wedde (JesperSchulz)]

Request changes: this does not currently demonstrate or deliver the claimed precision improvement. The target patch is an ordinary runtime codeunit, but both the evaluated BCApps upgrade instruction and al-upgrade-review relevance gate upgrade review to upgrade/install/schema/enum/migration surfaces; the changed article therefore cannot enter the deterministic worklist for this patch. Even if selected, the action contract evaluates ## Best Practice / ## Anti Pattern, while the new exclusion is only in ## Scope.

The offline result confirms the attribution gap: baseline and candidate both emitted zero findings, so F1=1 is vacuous and the gate is correctly marked inconclusive. Please reproduce the rejected upgrade-tag finding on the current baseline, then make the exclusion reachable through routing and normative guidance (or close/omit an inert knowledge change).