[Patrick-Erichsen]

Summary

Valid inline JSON can leave structural quotes on the shell parser’s conservative path, charging later prose to an unresolved command word. Prove and validate exact raw inline-code bodies before assigning quote ownership; keep payload contents available to security scanning.

 Markdown completeness
   validate whole JSON / JSON fences
+  validate canonically mapped inline JSON
   inspect structural quotes and raw payloads

Evidence

• Before: Unchanged Magneta 0.1.1 is 0/1 inspected; Whetstone is 32/34. Eight inline-JSON regression cases fail on the unchanged baseline.
  After: Magneta is 1/1; its full model scan completes in 30.60s with 5/5 successful analyzer calls. The real Go adapter accepts the unchanged raw report, retaining three findings and four reference caveats. Whetstone is 33/34; its TypeScript blocker remains incomplete.
• 22 focused controls, 1,301 installed runtime controls, and Go tests/vet/help pass. Native source controls: 1,833 passes and one unchanged-baseline expectation failure reproduced with native shfmt; no test or assertion was weakened.
• All 30 retained originals preserve every non-completeness finding, every original hash and all suppressions. Only one generated Whetstone coverage record is removed; the other 28 completeness/risk payloads match. All 138 reconstructed source/resource/dependency identities and 136 installed hashes match.
• Independent P0–P3 source review (four passes, complete modules) and runtime integration review are clear.

Refs openclaw/clawhub#4046. Production activation and original-upload recovery remain separate verification steps.

Merge Danger

Door: Two-way, through the scanner runtime pin.

Blast Radius: Scanner

Markdown completeness accounting changes. Dependencies, resource bounds, raw findings and verdict gates remain intact.

[clawsweeper]

🦞👀
ClawSweeper picked this up.

Pull request received. I will update this pull request when review starts.