[enhancement] Phase 0.4 leading-whitespace tolerance for CommonMark indent (verify follow-up from #53)

[kiki830621]

Problem

From verification of #53 (PR #58):
「Leading whitespace narrowing — ## Diagnosis (CommonMark legitimate 1-3 space indent) 從 substring true → 新 anchored regex false。Canonical IDD diagnosis comments 為 col-0,實務上不會踩到;但若 user 手動 paste 帶 indent 會 false-negative」
— Source: team:devils-advocate (HIGH self-rated) + codex (ACCEPTED RISK)

新 regex test("(?m)^## Diagnosis") 在嚴格 line-start anchor 下,完全排除 CommonMark spec 允許的 1-3 space leading indent ATX heading。

Type

enhancement (regex precision)

Priority

P3 (canonical IDD never hits; risk only on user manual paste)

Trigger scenario

GitHub comment body contains:
   ## Diagnosis     ← 3-space indent, CommonMark legitimate H2
   ...content...

Phase 0.4 detection:
- old contains("## Diagnosis") → true
- new test("(?m)^## Diagnosis") → false
- result: chain-shell falsely concludes "no diagnosis", prompts user re-run

Acceptance

• Decide: tolerate 1-3 space indent (CommonMark spec) OR keep strict col-0 (IDD canonical)
• If tolerate: regex → test("(?m)^[ \\t]{0,3}## Diagnosis"),smoke test 加 case 5 ( ## Diagnosis\n → true)
• If strict: doc add SKILL.md Step 0.4 footnote 「detection requires col-0 heading per IDD canonical convention」

Related

• fix: idd-all-chain Step 0.4 jq filter use anchored regex (avoid false-positive on quoted text) #53 (parent — anchored regex fix)
• cluster: chain follow-ups (#49 docs cross-link + #53 jq regex anchor) #58 (PR that introduced narrowing)

Source: surfaced during /idd-verify --pr 58 (cluster #49 #53), Step 5b triage. Master report: #58 (comment)