oh-my-customcode 0.62.5 → 0.63.0

package/README.md

@@ -21,17 +21,17 @@ npm install -g oh-my-customcode && cd your-project && omcustom init
 
 ---
 
-## What's New in v0.46.0
+## What's New in v0.62.5
 
 | Feature | Description |
 |---------|-------------|
-| **Rate Limit Monitoring** | Statusline now displays 5-hour rate limit usage with color-coded warnings (CC v2.1.80+) |
-| **Skill Effort Override** | Skills can set `effort` frontmatter to override model effort level at invocation time |
-| **Multi-project Web UI** | `omcustom serve` supports multi-project management with sidebar project selector |
-| **Batch Update UI** | Web dashboard supports visual project update status and batch updates |
-| **CC v2.1.72~v2.1.81 Compatibility** | Rate limits statusline, skill effort frontmatter, settings-based plugin source; `--bare` flag (harness disabled in bare mode), `--channels` permission relay (no changes required) |
-| **SDD Workflow** | Spec-Driven Development with `sdd/` folder hierarchy and planning-first gates |
-| **Ambiguity Gate** | Pre-routing clarity scoring and clarification questions |
+| **D3 Dependency Graph** | Interactive force-directed graph visualization at `/graph` — zoom, pan, drag, search, type filters |
+| **Playwright E2E Tests** | 11 accessibility tests with axe-core audit, `.pw.ts` extension for test isolation |
+| **Graph Accessibility** | WCAG keyboard navigation, aria-live announcements, skip link, focus-visible, reduced-motion support |
+| **CI Lockfile-Sync Gate** | New CI job validates bun.lockb consistency before lint/test |
+| **Token Optimization** | HTML comment technique reduces CLAUDE.md from 550→286 lines (48% reduction) |
+| **Workflow Engine** | YAML-defined workflow pipelines with `auto-dev` 7-step release batch |
+| **CC v2.1.83–v2.1.85 Compat** | Conditional hook `if` field, CwdChanged/FileChanged events, managed-settings.d |
 
 ---
 
@@ -131,7 +131,7 @@ Agent(arch-documenter):haiku      ┘
 |----------|-------|--------|
 | Languages | 6 | lang-golang, lang-python, lang-rust, lang-kotlin, lang-typescript, lang-java21 |
 | Backend | 6 | be-fastapi, be-springboot, be-go-backend, be-express, be-nestjs, be-django |
-| Frontend | 4 | fe-vercel, fe-vuejs, fe-svelte, fe-flutter |
+| Frontend | 5 | fe-vercel, fe-vuejs, fe-svelte, fe-flutter, fe-design |
 | Data Engineering | 6 | de-airflow, de-dbt, de-spark, de-kafka, de-snowflake, de-pipeline |
 | Database | 4 | db-supabase, db-postgres, db-redis, db-alembic |
 | Tooling | 3 | tool-npm, tool-optimizer, tool-bun |

package/dist/cli/index.js

@@ -9325,7 +9325,7 @@ var init_package = __esm(() => {
     workspaces: [
       "packages/*"
     ],
-    version: "0.62.5",
+    version: "0.63.0",
     description: "Batteries-included agent harness for Claude Code",
     type: "module",
     bin: {

package/dist/index.js

@@ -1674,7 +1674,7 @@ var package_default = {
   workspaces: [
     "packages/*"
   ],
-  version: "0.62.5",
+  version: "0.63.0",
   description: "Batteries-included agent harness for Claude Code",
   type: "module",
   bin: {

package/package.json

@@ -3,7 +3,7 @@
   "workspaces": [
     "packages/*"
   ],
-  "version": "0.62.5",
+  "version": "0.63.0",
   "description": "Batteries-included agent harness for Claude Code",
   "type": "module",
   "bin": {

package/templates/.claude/rules/SHOULD-ecomode.md

@@ -36,6 +36,35 @@ Ecomode: `[lang-golang-expert] ✓ src/main.go reviewed: 1 naming issue (handle_
 
 Disable with: "ecomode off", "verbose mode", or "show full details".
 
+## Input Context Pruning
+
+Active removal of irrelevant retrieved content from agent context. Complements output compression by managing the input side of token budget.
+
+> **Terminology**: "Input Context Pruning" (R013) manages retrieved chunks during a task. "Memory Pruning" (R011) manages behavioral memory across sessions. These are distinct concepts.
+
+### Pruning Triggers
+
+| Trigger | Condition | Action |
+|---------|-----------|--------|
+| Search overflow | Retrieved chunks > 10 | Retain top-K by relevance, prune rest |
+| Context pressure | Context usage > 50% | Summarize oldest/lowest-relevance chunks |
+| Multi-hop intermediate | Between retrieval hops | Replace previous hop raw results with summary |
+
+### Pruning Strategy
+
+| Strategy | When | Behavior |
+|----------|------|----------|
+| **Retain** | Directly relevant code/docs | Keep as-is |
+| **Summarize** | Background context, prior hop results | Replace with 1-2 line summary |
+| **Drop** | Search noise, duplicates, already-reflected info | Remove entirely |
+
+### Rules
+
+- Pruning is irreversible — generate summary BEFORE dropping original
+- Prune at document/chunk level, not mid-sentence
+- When in doubt, Summarize rather than Drop
+- Track pruning decisions: `[Pruned] {N} chunks → {M} retained, {K} summarized, {J} dropped`
+
 ## Context Budget Management
 
 Task-type-aware context thresholds that trigger ecomode earlier for context-heavy operations.

package/templates/.claude/skills/memory-recall/SKILL.md

@@ -58,6 +58,24 @@ avoid:
   - Special characters or complex syntax
 ```
 
+## Retrieval Strategy
+
+### Recall-Precision Tradeoff
+
+Default bias: **recall > precision** — it is easier to filter out irrelevant results (false positives) than to recover missed information (false negatives).
+
+| Task Type | Bias | Recommended --limit | Rationale |
+|-----------|------|--------------------:|-----------|
+| Debugging / Investigation | High recall (16:1) | 10-15 | Cast wide net, prune later |
+| Decision reference | Balanced | 5 (default) | Moderate breadth with manageable noise |
+| Specific fact lookup | High precision | 3 | Narrow, targeted retrieval |
+
+### Guidelines
+
+- **Over-retrieve, then filter**: When uncertain, request more results and discard irrelevant ones in post-processing
+- **Narrow progressively**: Start broad, refine query only if results are noisy — avoid starting too narrow
+- **Combine temporal + semantic**: Add date filters (`--date`) to semantic queries for better precision without sacrificing recall
+
 ## Output Format
 
 ### Basic Search

package/templates/.claude/skills/research/SKILL.md

@@ -228,6 +228,33 @@ Convergence expected by round 3. Hard stop at round 30.
 | Ecomode | Auto-activate for team result aggregation (R013) |
 | Intent display | Show research plan before execution (R015) |
 
+## Retrieval-Reasoning Separation
+
+Retrieval and reasoning are distinct cognitive operations that benefit from explicit role separation. Mixing them in a single agent degrades both: retrieval becomes biased by premature conclusions, and reasoning gets polluted by search noise.
+
+### Principle
+
+| Role | Phase | Model | Responsibility |
+|------|-------|-------|----------------|
+| Retriever | Phase 1 | sonnet (fast, broad) | Gather, catalog, enumerate — no judgment |
+| Reasoner | Phase 2-3 | opus (deep, precise) | Verify, synthesize, judge — no new retrieval |
+
+### Why Separate
+
+- **Retrieval bias**: A reasoning agent searching for evidence tends to confirm existing hypotheses (confirmation bias)
+- **Context pollution**: Raw search results mixed with analysis obscure both
+- **Cost efficiency**: Retrieval needs speed and breadth (cheaper model); reasoning needs depth (capable model)
+- **Debuggability**: When results are wrong, separated roles make it clear whether the problem was bad retrieval or bad reasoning
+
+### Application in Research Workflow
+
+| Phase | Role | Separation Rule |
+|-------|------|-----------------|
+| Phase 1 (10 teams) | Retriever | Teams gather and catalog only. No ADOPT/AVOID judgments. |
+| Phase 2 (Verification) | Reasoner | Verifiers challenge claims using Phase 1 data. No new searches. |
+| Phase 3 (Synthesis) | Reasoner | Synthesizer produces taxonomy from verified findings only. |
+| Phase 4 (Output) | Reporter | Formats and persists. No new analysis. |
+
 ## Model Selection
 
 | Phase | Model | Rationale |

package/templates/manifest.json

@@ -1,5 +1,5 @@
 {
-  "version": "0.62.5",
+  "version": "0.63.0",
   "lastUpdated": "2026-03-24T00:00:00.000Z",
   "components": [
     {