Ecosystem Update — 2026-04-19

April 19, 2026 · generated by the ecosystem-update Claude Skill

Dry-run — no changes applied to settings.json, agents, or skills. State file not updated.

TL;DR

Very quiet cycle: all 4 "latest additions" from awesome-claude-code and all major Boris/shanraisshan items already in seen_items (148 prior items).
Net-new discoverables are mostly built-in CLI flags and commands (--bare, --add-dir, --agent, /sandbox, /voice) — knowledge-level, not config-level. No Quick Wins today.
Tier 2 (arxiv) surfaced 8 new April-2026 papers — notable: AgentForge (execution-grounded multi-agent with mandatory Docker verification), ORBIT (dependency-aware translation graph), and Three Roles One Model (role scaffolding at inference time).

Quick Wins

None. No MISSING candidate passes the Quick-Win gate (Priority ≥ 2.0 AND hard-limit-safe — no new files, no new hook scripts, frontmatter-only for agents). The genuinely missing items are either (a) built-in CLI features already callable without config changes, or (b) Build-Queue items that require new scripts.

Build Queue

Prompt-hook guardrail (hook) — claude-code-new-features-early-2026 — A PreToolUse hook that calls a cheap LLM to validate a dangerous tool call before it executes. Cost: pennies per check. Would complement pre_tool_guard.py (which is a deterministic rule checker) with a semantic layer. Requires a new script at ~/.claude/bin/prompt_hook_validator.py. Impact 3 / Effort 3 / Priority 1.0 / Align Y.
/careful and /freeze skill blocks (skill-pattern) — claude-code-best-practice — On-demand permission gates embeddable in SKILL.md to restrict destructive commands and scope edits to named directories during that skill's execution. Useful for the refactor, audit, and security-audit skills which touch many files. Impact 2 / Effort 2 / Priority 1.0 / Align Y.
/code-review multi-agent PR analysis command (skill) — awesome-claude-code-toolkit — Multi-agent PR analyzer (bugs + security + regressions via parallel agents). Partially redundant with existing reviewer + security-review + codex-branch trio, but packaged as a single /code-review <PR> entry point. Impact 2 / Effort 2 / Priority 1.0 / Align Y — marginal, consider only if current reviewer workflow feels under-integrated.

CLI Awareness (not config changes)

These are built-in Claude Code features not previously in seen_items. No install step — they are callable today. Flagged for user awareness only.

Feature	What it does	When to use
`claude --bare`	Skips CLAUDE.md / settings.json / MCP discovery	~10× faster startup in non-interactive SDK scripts
`claude --add-dir <path>`	Grants access to an extra repo at launch	Cross-repo work without moving files
`claude --agent=<name>`	Launch bound to a specific agent (e.g. `--agent=explorer`)	Restricted-tool read-only sessions
`/sandbox`	Opens the OSS sandbox runtime for file/network isolation	Reduces permission prompts safely on untrusted tasks
`/voice`	Voice dictation (hold space in CLI)	Claimed ~3× typing speed; shanraisshan reports ~10× with Wispr Flow
Recap auto-inject	Short auto-summaries in long sessions, toggle via `/config`	Disable if summaries feel noisy

Research

April 2026 arxiv hits relevant to the current multi-agent setup. All go to Research bucket — not actionable until validated.

AgentForge: Execution-Grounded Multi-Agent LLM Framework — Planner/Coder/Tester/Debugger coordination with mandatory Docker sandbox verification between turns. Relevant: current validator agent is stdlib-only; a container-verified stage could strengthen R3/R4 acceptance gates.
ORBIT: Guided Agentic Orchestration for C-to-Rust Transpilation — Dependency-aware translation graph with 91.7% success on >1000 LoC. Relevant: general pattern for dependency-ordered slice execution in auto-runtime.
Three Roles, One Model: Role Orchestration at Inference Time — Role scaffolding at inference closes the small/large-model gap without training. Relevant: current planner/worker/reviewer/validator split is exactly this pattern — paper gives empirical priors on where the gains concentrate.
λ_A: Typed Lambda Calculus for LLM Agent Composition — Formalizes type safety and termination for agent composition, identifies structural incompleteness in real configs. Relevant: possible framework for auditing route_manifest.json composition rules.
FM-Agent: Hoare-Style Reasoning at 143K LoC — Top-down compositional function-spec derivation from caller expectations. Relevant to large-codebase R4 work.
COEVO: Joint Functional Correctness + PPA Optimization — Single evolutionary loop with fine-grained testbench scoring. Relevant to verifier convergence design.
WebXSkill: Executable Skill Learning for Web Agents — Parameterized programs paired with NL guidance, dual-mode (direct exec + agent-adapted). Relevant to the Skill tool design philosophy.
HWE-Bench: LLM Hardware Bug Repair — Benchmark showing fault localization and cross-artifact coordination as dominant failure modes — applies to any multi-file agent task, not just hardware.

Already Have

claude-pace, clawd-on-desk, ccxray, encyclopedia-agentic-coding-patterns, postcompact-hook, worktreecreate-remove-hooks, auto-mode, simplify-command, batch-command, go-composite-skill, btw-side-queries, imessage-plugin, rc-remote-control, rename-session, permission-mode-plan, important-if-conditional-tag, claude-rules-subdirectory-modularization, opus-permission-routing-tool, instinct-scoring-agentshield, claude-code-auto-compact-window-env, xhigh-opus-47-default-effort, skills-janitor, reviewgrounder-rubric-reviewer, autogenesis-paper, anypoc-paper, at-claude-github-action, tdd-guard-installed, context-fork-pattern, isolation-worktree-worker / -reviewer / -planner / -python-reviewer / -typescript-reviewer.

Rejected

claude --name / auto-naming — session naming; already covered by rename-session in seen_items and not a config change.
/rewind / /compact vs /clear — built-in context controls, no install needed.
BigQuery integration / Slack MCP — workflow-specific, not general-purpose; alignment fail (no concrete recurring need in current setup).
Chrome Extension for frontend verification — browser-tier feature, parallel to existing Playwright MCP (already installed).
Comprehensive toolkit drops (rohitg00, travisjneuman, vinicius91carvalho) — bulk catalogs; extract individual items on demand rather than installing wholesale. Alignment fail at the catalog level.
PostCompact hook as Quick Win — in seen_items but still not wired; hard-limit forbids registering a hook whose script does not yet exist. Move to Build Queue next cycle if the script is authored first.

Sources checked: awesome-claude-code, howborisusesclaudecode.com, claude-code-best-practice, arxiv LLM agent query, WebSearch GitHub 2026. Tier 2 fetched: yes (arxiv, last fetch 2026-04-16). Tier 3 fetched: no (last 2026-04-13, within 7-day window). Run at: 2026-04-19 (dry-run). Prior seen_items count: 148. New items this run: 0 added to state (dry-run preserves state for real run).