style-review: calibrate analyzer-redundant rules to info; keep correctness bugs out of style scope (#95)

Online-eval data shows the style leaf is the largest source of dismissed
findings. Two causes:

- Mechanical, analyzer-enforced conventions (this-keyword AA0248, label
  suffixes AA0074, missing ToolTip, label scope) fire at gating severity
  and duplicate what CodeCop/AppSourceCop already report. Calibrate them
  to info so a severity-gating consumer drops the redundant noise.
- Correctness/logic/data-integrity defects get reframed as style
  conventions and emitted here. Sharpen the scope boundary: such defects
  belong to the relevant domain leaf, or to al-code-review's cross-cutting
  agent channel when no knowledge file covers them, never to this leaf.

Co-authored-by: wenjiefan <wenjiefan@microsoft.com>

microsoft/skills/review/al-style-review.md

@@ -53,13 +53,15 @@ When the post-conflict worklist is empty because no applicable style knowledge e
 
 For each worklist entry, evaluate the diff against the file's `## Best Practice` and `## Anti Pattern` sections. Style findings rarely reach `blocker` — reserve it for cases where the knowledge file documents a platform-level requirement (for example, API page property constraints the OData runtime rejects). Most style findings are `minor` or `info`; egregious misuse (`Error` with pre-built Text losing translation and telemetry classification) may reach `major`.
 
+Severity calibration — a formal analyzer already flags the mechanical presence/naming conventions (the `this` keyword AA0248, approved label suffixes AA0074, a missing `ToolTip`, a `Label` declared at local instead of object scope, required parentheses). On those, BCQuality's value is the *explanation* of why the rule exists, not a second gate; emit them at `info` so a consumer that gates on severity does not re-flag what CodeCop/AppSourceCop already reports. Reserve `minor` for style issues with concrete downstream impact the analyzer does not catch — lost translation or telemetry classification from a string-built `Error`, an `OptionCaption` that does not match its `OptionMembers`, a misleading named invocation. This keeps the domain's default output advisory and prevents analyzer-redundant noise from competing with substantive review.
+
 Set `confidence` to:
 
 - `high` when the detection is based on an unambiguous pattern match.
 - `medium` when detection relies on heuristics or when any frontmatter dimension was `unknown`.
 - `low` when the finding is an advisory derived only from applicability.
 
-After evaluating each worklist entry, also consider whether the diff exhibits a style defect the agent recognises from its general AL knowledge that no knowledge file in the worklist covers. Such candidates are agent findings within this skill's domain — emit them with `references: []`, an `id` slug prefixed with `agent:`, `confidence` capped at `medium`, `severity` capped at `minor` (agent findings are advisory and non-gating), and a `message` that is self-contained (describing both the issue and a concrete recommendation, since there is no knowledge-file footer for the consumer to fall back on). Hold every candidate to the precision bar in `skills/do.md` (*Agent findings*): emit only a clear, widely-accepted AL style violation with a concrete basis a knowledgeable BC reviewer would agree on — steelman it first and drop personal preference, speculation, and any single defensible formatting choice among several; when in doubt, omit. The scope is strictly style; defects outside this domain belong to other leaves and MUST NOT be emitted here. Before emitting, check the worklist for a knowledge file that matches the candidate — if one exists, upgrade the candidate to a knowledge-backed finding instead. See `skills/do.md` for the full contract.
+After evaluating each worklist entry, also consider whether the diff exhibits a style defect the agent recognises from its general AL knowledge that no knowledge file in the worklist covers. Such candidates are agent findings within this skill's domain — emit them with `references: []`, an `id` slug prefixed with `agent:`, `confidence` capped at `medium`, `severity` capped at `minor` (agent findings are advisory and non-gating), and a `message` that is self-contained (describing both the issue and a concrete recommendation, since there is no knowledge-file footer for the consumer to fall back on). Hold every candidate to the precision bar in `skills/do.md` (*Agent findings*): emit only a clear, widely-accepted AL style violation with a concrete basis a knowledgeable BC reviewer would agree on — steelman it first and drop personal preference, speculation, and any single defensible formatting choice among several; when in doubt, omit. The scope is strictly style — naming, labelling, formatting, and analyzer-adjacent conventions. A correctness, logic, data-integrity, or contract defect is NOT a style finding even when it can be reworded as a convention: a method that mutates a shared `Record`'s filters, an unfiltered `DeleteAll`, a violated interface contract, or a wrong boolean guard are behavioural defects, not conventions — do not emit them here under a style framing. If a specific domain leaf covers the concern (performance, security, error-handling, …) it belongs there; if no knowledge file in any domain covers it, it belongs to the `al-code-review` super-skill's cross-cutting self-review agent channel (`from-sub-skill: "agent"`, `severity` capped at `minor`), not to this leaf. A reliable test: if you cannot cite a style `## Best Practice`/`## Anti Pattern` for the concern, it is very likely not a style finding. Before emitting, check the worklist for a knowledge file that matches the candidate — if one exists, upgrade the candidate to a knowledge-backed finding instead. See `skills/do.md` for the full contract.
 
 For every emitted finding, decide whether the fix is mechanical. A fix is mechanical when it is small, local, and unambiguous from the diff context (for example: delete unreachable lines; replace `Count() > 0` with `not IsEmpty()`; move a local `Label` to object scope; add a missing `ToolTip`, `OptionCaption`, or `DataClassification`; replace a string-concatenated `Error` with a Label-backed call; change an over-broad permission token; or add an obvious `else`/guard branch). For mechanical findings, emit `findings[].suggested-code` with the literal replacement for the source lines indicated by `location`. The payload must be a verbatim replacement — no diff markers, no fences, no commentary — that the consumer can render as a one-click suggestion. When a `.good.al` companion exists and the diff context matches the `.bad.al` shape, adapt the `.good.al` replacement into `suggested-code`.
 
@@ -82,13 +84,13 @@ Output conforms to the DO output contract. A populated example:
   "skill": { "id": "al-style-review", "version": 1 },
   "outcome": "completed",
   "summary": {
-    "counts": { "blocker": 0, "major": 0, "minor": 1, "info": 0 },
+    "counts": { "blocker": 0, "major": 0, "minor": 0, "info": 1 },
     "coverage": { "worklist-size": 1, "items-evaluated": 1 }
   },
   "findings": [
     {
-      "id": "microsoft/knowledge/style/label-suffix-approved-list.md",
+      "id": "microsoft/knowledge/style/apply-approved-label-suffixes.md",
-      "severity": "minor",
+      "severity": "info",
       "message": "A Label named Text000 has no approved suffix (Msg/Err/Qst/Tok/Lbl/Txt). Per the referenced CodeCop AA0074 guidance, every Label and TextConst carries a suffix indicating its consuming call.",
       "location": {
         "file": "src/Sales/PostingRoutines.Codeunit.al",