oh-my-customcode 0.173.0 → 0.175.0

package/dist/cli/index.js

@@ -241,7 +241,7 @@ var init_package = __esm(() => {
     workspaces: [
       "packages/*"
     ],
-    version: "0.173.0",
+    version: "0.175.0",
     description: "Batteries-included agent harness for Claude Code",
     type: "module",
     bin: {
@@ -289,7 +289,7 @@ var init_package = __esm(() => {
       yaml: "^2.8.2"
     },
     devDependencies: {
-      "@anthropic-ai/sdk": "^0.100.1",
+      "@anthropic-ai/sdk": "^0.102.0",
       "@biomejs/biome": "^2.3.12",
       "@types/bun": "^1.3.6",
       "@types/js-yaml": "^4.0.9",

package/dist/index.js

@@ -2031,7 +2031,7 @@ var package_default = {
   workspaces: [
     "packages/*"
   ],
-  version: "0.173.0",
+  version: "0.175.0",
   description: "Batteries-included agent harness for Claude Code",
   type: "module",
   bin: {
@@ -2079,7 +2079,7 @@ var package_default = {
     yaml: "^2.8.2"
   },
   devDependencies: {
-    "@anthropic-ai/sdk": "^0.100.1",
+    "@anthropic-ai/sdk": "^0.102.0",
     "@biomejs/biome": "^2.3.12",
     "@types/bun": "^1.3.6",
     "@types/js-yaml": "^4.0.9",

package/package.json

@@ -3,7 +3,7 @@
   "workspaces": [
     "packages/*"
   ],
-  "version": "0.173.0",
+  "version": "0.175.0",
   "description": "Batteries-included agent harness for Claude Code",
   "type": "module",
   "bin": {
@@ -51,7 +51,7 @@
     "yaml": "^2.8.2"
   },
   "devDependencies": {
-    "@anthropic-ai/sdk": "^0.100.1",
+    "@anthropic-ai/sdk": "^0.102.0",
     "@biomejs/biome": "^2.3.12",
     "@types/bun": "^1.3.6",
     "@types/js-yaml": "^4.0.9",

package/templates/.claude/rules/MUST-agent-design.md

@@ -27,6 +27,16 @@ tools: [Read, Write, ...]  # Allowed tools
 
 Extended context suffix: `[1m]` (e.g., `claude-opus-4-6[1m]`) — enables 1M token context window.
 
+### Fallback Models (CC v2.1.166+)
+
+> **v2.1.166+**: The `fallbackModel` setting configures up to three fallback models tried in order when the primary model is overloaded or unavailable. `--fallback-model` now also applies to interactive sessions. CC additionally retries a turn once on the fallback model when the API rejects an unexpected non-retryable error (auth, rate-limit, request-size, and transport errors still surface immediately).
+
+This is a settings-level resilience mechanism, distinct from the per-agent `model:` frontmatter. It complements the `model-escalation` skill (outcome-based escalation) by handling availability/overload failover at the platform level.
+
+### Thinking Toggle (CC v2.1.166+)
+
+> **v2.1.166+**: `MAX_THINKING_TOKENS=0`, `--thinking disabled`, and the per-model thinking toggle disable thinking on models that think by default via the Claude API (3rd-party providers unchanged). Relevant when an agent's `effort` is low and thinking overhead is undesirable.
+
 ### Optional Frontmatter
 
 Key optional fields: `memory`, `effort`, `skills`, `soul`, `isolation`, `background`, `maxTurns`, `maxTokens`, `mcpServers`, `hooks`, `permissionMode`, `disallowedTools`, `limitations`, `domain`, `disableSkillShellExecution`. Supported since CC v2.1.63+. See full optional frontmatter via Read tool.

package/templates/.claude/rules/MUST-agent-teams.md

@@ -36,6 +36,18 @@ Available when `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` or TeamCreate/SendMessag
 
 These are distinct mechanisms. Agent Teams `SendMessage` requires `TeamCreate` and operates within a single Claude Code session. claude-peers-mcp `send_message` operates across separate Claude Code processes via a localhost broker.
 
+### Cross-Session Relay Authority Hardening (CC v2.1.166+)
+
+> **v2.1.166+**: Messages relayed via `SendMessage` from other Claude sessions no longer carry user authority — receivers refuse relayed permission requests, and auto mode blocks them. A relayed message cannot escalate privilege on the receiving session.
+
+| Aspect | Behavior (v2.1.166+) |
+|--------|---------------------|
+| Relayed permission request | Refused by receiver |
+| Auto mode + relayed request | Blocked |
+| User authority across relay | Not propagated |
+
+This hardens cross-session coordination (claude-peers-mcp `send_message`, see Scope table above) against privilege escalation — a relayed message from session A cannot grant session B permissions the user did not authorize on B. Aligns with R001 (credential/privileged-scope guardrails) and R010 (out-of-scope privileged chaining). Intra-session Agent Teams `SendMessage` between peers in the same session is unaffected.
+
 ## Self-Check (Before Agent Tool)
 
 Before using Agent tool for 2+ agent tasks, complete this check:

package/templates/.claude/rules/MUST-parallel-execution.md

@@ -50,6 +50,17 @@ Before writing/editing multiple files:
 
 > **Token threshold heuristic**: When a delegated agent prompt exceeds ~5000 tokens or spans 3+ unrelated domains, decompose by domain and spawn parallel agents. See R018 for Agent Teams criteria when review cycles are needed. Reference: #1085.
 
+### LLM Batch Output Token Budget
+
+The giant-prompt heuristic above governs INPUT tokens. The symmetric OUTPUT-side rule: when a single LLM call processes N items (scoring/classifying/extracting) and must emit structured output (e.g. JSON) per item, pre-compute the output budget = N × per-item output tokens BEFORE the call. Exceeding `max_tokens` truncates the response mid-structure → silent parse failure (the call "succeeds" but JSON.parse throws).
+
+| Anti-pattern | Required |
+|--------------|----------|
+| Single batch call over a variable-size list with a fixed small max_tokens | Chunk into ≤40-item batches; constrain per-item output length (e.g. reason ≤10 words); raise max_tokens to fit one chunk |
+| Raising max_tokens alone | Insufficient — defers the failure as the list grows. Chunking is the invariant fix. |
+
+Reference: #1320 (fix), #1321 (session 113 retrospective 찐빠 #1), `feedback_llm_batch_truncation.md`.
+
 <!-- DETAIL: Full violation examples (4 pairs)
 ❌ WRONG: Writing files one by one
    Write(file1.kt) → Write(file2.kt) → Write(file3.kt) → Write(file4.kt)

package/templates/.claude/rules/MUST-permissions.md

@@ -33,6 +33,18 @@ Approve?
 
 On insufficient permission: do not attempt, notify user, suggest alternative.
 
+## Deny Rule Glob Patterns (CC v2.1.166+)
+
+> **v2.1.166+**: Deny rules support glob patterns in the tool-name position — `"*"` in a deny rule denies all tools. Allow rules reject non-MCP globs (only MCP tool-name globs are accepted in the allow position). Unknown tool names in deny rules emit a startup warning.
+
+| Position | Glob support |
+|----------|-------------|
+| Deny rule tool-name | Yes — `"*"` denies all tools |
+| Allow rule tool-name | MCP globs only; non-MCP globs rejected |
+| Unknown tool in deny rule | Startup warning |
+
+Use a `"*"` deny rule in `settings.json` to enforce a deny-by-default posture, then add specific allow rules. Complements the Tier-based policy above — settings.json deny rules are evaluated by the CC platform, independent of the advisory tier table.
+
 ## Agent Tool Permission Mode
 
 When spawning subagents via the Agent tool, always pass `mode: "bypassPermissions"` explicitly. The Agent tool's default mode is `acceptEdits`, which **overrides** the agent frontmatter `permissionMode` field.

package/templates/.claude/rules/MUST-tool-identification.md

@@ -81,6 +81,21 @@ matches the spawn announcement:
   [2] lang-python-expert:sonnet → Python code review
 ```
 
+## Tier-3 Interaction Tool Prefix (MANDATORY)
+
+R008 "every tool call" applies to Tier-3 interaction tools too — NOT only file/exec tools. Applying the `[agent][model] → Tool:` prefix to Agent/Bash/Read while omitting it on `AskUserQuestion`, `TodoWrite`, `EnterPlanMode`, etc. is a violation.
+
+| Tool | R008 prefix required? |
+|------|----------------------|
+| AskUserQuestion | YES — `[agent][model] → Tool: AskUserQuestion` before the call |
+| TodoWrite | YES |
+| EnterPlanMode / ExitPlanMode | YES |
+| Skill | NO separate R008 prefix — identified via R007 `claude → {skill-name}` integrated header instead |
+
+Skill invocation is the one exception: it is identified through the R007 integrated identification block (`┌─ Agent: claude → {skill-name}`), not a standalone R008 tool prefix.
+
+Reference issue: #1321 (session 113 retrospective, 찐빠 #2 — AskUserQuestion prefix omitted twice).
+
 ## Example
 
 ```

package/templates/manifest.json

@@ -1,5 +1,5 @@
 {
-  "version": "0.173.0",
+  "version": "0.175.0",
   "lastUpdated": "2026-05-20T00:00:00.000Z",
   "omcustomMinClaudeCode": "2.1.121",
   "omcustomMinClaudeCodeReason": "Sensitive-path direct Write/Edit on .claude/** under bypassPermissions (R010 deprecation, #1101)",