chore: Switch finding-id format to mi-*/I-* to avoid Windows case-insensitive collisions

Author: hanatyan128

.claude/rules/commit.md

@@ -14,7 +14,7 @@ Rules that apply to every commit made in this repository.
 ## Content to avoid
 
 - Do NOT include workflow-specific labels such as "Review response", "レビュー対応", or similar. The commit message should describe **what the code does now**, not the process that produced it.
-- Do NOT include finding IDs from review documents (e.g. `C-1`, `M-1`, `m-1`, `I-1`). These are internal to the review workflow and have no meaning outside of it.
+- Do NOT include finding IDs from review documents (e.g. `C-1`, `M-1`, `mi-1`, `I-1`). These are internal to the review workflow and have no meaning outside of it.
 
 ## Trailers
 

.claude/skills/parallel-review/SKILL.md

@@ -136,7 +136,7 @@ Aggregation procedure:
 1. Read each reviewer file.
 2. Deduplicate — when multiple reviewers point out the same location with the same intent, merge them into one entry and record the originating reviewers.
 3. Group by severity (Critical → Major → Minor → Info).
-4. Within each group, assign finding-ids (C-1, C-2, ..., M-1, ..., m-1, ..., i-1).
+4. Within each group, assign finding-ids (Critical: C-1, C-2, ...; Major: M-1, M-2, ...; Minor: mi-1, mi-2, ...; Info: I-1, I-2, ...).
 5. Read `.claude/skills/parallel-review/templates/review-doc.md` to grasp the template skeleton, then Write to {final_doc_path} following the format rules in this SKILL § Format Rules.
 
 Return value: {doc_path, findings_total, severity_counts: {critical, major, minor, info}, duplicates_merged}

.claude/skills/review-respond/SKILL.md

@@ -128,7 +128,7 @@ Launch procedure:
 ```
 You are responsible for parsing the review document. Read {document_path} and Write the metadata for each finding to {tmp_dir}/findings.json.
 
-Extraction targets: Critical / Major / Minor sections (skip Info). For each finding extract id (C-1, M-1, m-1, etc.) / severity / location / description (body up to the markers) / current_meta (current values of triage / estimate / status / verification; when the same field appears multiple times, take the last value).
+Extraction targets: Critical / Major / Minor sections (skip Info). For each finding extract id (C-1, M-1, mi-1, etc.) / severity / location / description (body up to the markers) / current_meta (current values of triage / estimate / status / verification; when the same field appears multiple times, take the last value).
 
 stage classification (based on current_meta):
 - markers empty → pending_triage