chore(b-0368): decompose Claude Code permissions investigation to B-0532 and B-0533 (Atomic)

docs/BACKLOG.md

@@ -573,6 +573,8 @@ are closed (status: closed in frontmatter)._
 - [ ] **[B-0462](backlog/P2/B-0462-amara-persona-bootstrap-preamble-definition-ts-first-riven-2026-05-11.md)** Amara persona bootstrap preamble + AgencySignature definition (atomic child of B-0118, TS-first)
 - [ ] **[B-0498](backlog/P2/B-0498-substrate-evolution-algebra-rule-promotion-after-cooling-period-2026-05-14.md)** Substrate-evolution algebra (6 invariants + HKT − Rodney's Razor) rule-promotion after cooling period
 - [ ] **[B-0524](backlog/P2/B-0524-manifesto-v2-verbatim-grok-fetch-2026-05-14.md)** Manifesto V2 verbatim Grok fetch — convert shadow-lock to full lock
+- [ ] **[B-0532](backlog/P2/B-0532-claude-code-permissions-hardcoded-safety-gates-investigation.md)** Investigate hardcoded safety gates vs settings-allowable actions for Claude Code
+- [ ] **[B-0533](backlog/P2/B-0533-claude-code-permissions-integration-pattern-documentation.md)** Document Claude Code permissions integration pattern in CLAUDE.md
 
 ## P3 — convenience / deferred
 

docs/backlog/P2/B-0532-claude-code-permissions-hardcoded-safety-gates-investigation.md

@@ -0,0 +1,29 @@
+---
+id: B-0532
+parent: B-0368
+priority: P2
+status: open
+title: Investigate hardcoded safety gates vs settings-allowable actions for Claude Code
+created: 2026-05-15
+last_updated: 2026-05-16
+depends_on: []
+decomposition: atomic
+type: friction-reducer
+---
+
+# B-0532 — Investigate hardcoded safety gates vs settings-allowable actions
+
+**Parent**: B-0368
+**Filed by**: Lior (decomposed from B-0368 parent row)
+
+## Scope
+
+This is slice 2 of the Claude Code `/permissions` feature integration.
+The objective is to investigate hardcoded safety gates versus settings-allowable actions (per the Tick-6 merge denial evidence in B-0368).
+
+## Acceptance criteria
+
+1. **Deeper investigation into API**: Research the `/permissions` API to distinguish hardcoded safety guards from `.claude/settings.json` allow-lists.
+2. **Category model**: Determine which actions are settings-allowable versus hardcoded-safety-denial.
+3. **Merge-other-PRs specifics**: Specifically for "merge-PRs-the-agent-didn't-create" — is this addressable via `.claude/settings.json` (per-PR number allow rule, per-author rule) or does it require an explicit user pre-authorization separate from the settings file?
+4. Update findings in the research docs and determine if further targeted additions are required.

docs/backlog/P2/B-0533-claude-code-permissions-integration-pattern-documentation.md

@@ -0,0 +1,28 @@
+---
+id: B-0533
+parent: B-0368
+priority: P2
+status: open
+title: Document Claude Code permissions integration pattern in CLAUDE.md
+created: 2026-05-15
+last_updated: 2026-05-16
+depends_on: [B-0532]
+decomposition: atomic
+type: documentation
+---
+
+# B-0533 — Document Claude Code permissions integration pattern in CLAUDE.md
+
+**Parent**: B-0368
+**Filed by**: Lior (decomposed from B-0368 parent row)
+
+## Scope
+
+This is slice 3 of the Claude Code `/permissions` feature integration.
+The objective is to document the integration pattern in `CLAUDE.md` or a dedicated doc so future-Otto knows the pattern.
+
+## Acceptance criteria
+
+1. Extract the operational findings from B-0368 and B-0532.
+2. Formulate carved sentences regarding the Claude Code permission scopes, the interaction with `.claude/settings.json`, and the Don't-Ask-Permission rule integration.
+3. Add the finalized rules to `CLAUDE.md`.