Experiment 01 · recipes

Workflow recipes, mined from the corpus.

148 concrete multi-feature workflow hypotheses across 55 Claude Code versions. Each one names the primitives it combines and the thesis for why the combination unlocks something neither does alone. Changelog-derived and mostly untested, published as raw material for anyone building agent systems.

148recipes
55versions
9validated
2.1.0 Rapid Skill Development Loop VALIDATED ✅
#1 hot-reload + #2 context:fork

Hot-reload lets you iterate instantly. Fork isolates each test run so failed experiments don't pollute your session. Together = fearless rapid iteration.

2.1.0 Parallel Research Synthesis VALIDATED ✅
#2 context:fork + #3 agent field

Fork creates isolated execution. Agent field routes to different models via agent types. Together = spawn multiple cheap/fast researchers, synthesize in expensive/smart orchestrator.

2.1.0 Self-Improving Skills VALIDATED ✅ 🤯
#1 hot-reload + #2 context:fork + skills that edit skills

Skills can edit files. Skills are files. Hot-reload picks up changes. A forked skill could analyze its own performance and modify itself.

2.1.0 Multi-Language Documentation Generator BLOCKED - language is global, not…
#2 context:fork + #4 language setting

Language setting changes Claude's response language. Fork isolates each language context. Together = parallel generation in multiple languages.

2.1.0 Skill Pipeline / Middleware Chain PARTIALLY VALIDATED
#1 hot-reload + hooks in skills (2.1.0 feature)

Skills can have hooks. Hooks can call other skills. Hot-reload means you can modify any stage of the pipeline live.

2.1.0 Audit Trail for Sensitive Operations VALIDATED ✅
#5 skill hooks + file logging

Hooks can log tool usage. Skill-scoped hooks = per-workflow audit without global noise.

2.1.0 Sandboxed Plugin Testing untested
#2 context:fork + tool restrictions

Fork + allowed-tools creates a sandbox. Test untrusted skills safely.

2.1.0 Context-Aware Skill Specialization untested
#1 hot-reload + skills that read project context

Skills can read files on invocation. Skill can modify itself based on what it reads. Hot-reload picks up changes.

2.1.0 Batch Processing with Checkpointing VALIDATED ✅
#2 context:fork + file-based state

Each fork is stateless. But forks can read/write files. Use files as checkpoint mechanism.

2.1.0 Model-Tiered Agent Pipeline VALIDATED
#3 agent field + #6 custom agents

Custom agents let you define model + tools. Skills can route to specific agents. Together = cost-optimized pipelines with capability matching.

2.1.0 Sandboxed Third-Party Skills VALIDATED
#6 custom agents + tool restrictions

Custom agents enforce tool access. Reference untrusted skill with restricted agent = sandboxed execution.

2.1.33 Persistent Multi-Agent Task Pipeline untested
TeammateIdle + TaskCompleted hook events + memory frontmatter (project scope) + Task(agent_type) spawning restrictions

Multi-agent workflows now have state continuity. Agents can persist context across delegations, react to peer lifecycle changes, and enforce spawning policies. This unlocks checkpoint-resumable pipelines where teammates hand off work while preserving shared state without re-initializing each step.

2.1.33 Inter-Session Teammate State Machine untested
memory frontmatter with project scope + TeammateIdle hook + remote sessions (VSCode support)

Agents can now maintain state across separate Claude Code sessions (VSCode remote, tmux, CLI). Hook events enable clean state transitions. Project-scoped memory acts as a durable handoff surface between independently-spawned teammate instances.

2.1.41 Headless CI/CD Credential Rotation Loop untested
claude auth login + claude auth status + claude auth logout

The auth CLI subcommands enable agents to manage credential lifecycle without interactive prompts, unlocking automated credential rotation, multi-account cycling, and safe session cleanup in orchestrated pipelines.

2.1.41 Observable Agent Performance Monitoring with OTel Speed Attribute untested
speed attribute (OTel) + existing trace/span instrumentation

The speed attribute exposes fast-mode execution in OTel spans, enabling agents to self-monitor performance and make runtime decisions about trading speed for quality (e.g., falling back to thorough mode if fast-mode latency regresses).

2.1.45 Rate-limit-aware agent work queuing untested
SDKRateLimitInfo / SDKRateLimitEvent + Claude Sonnet 4.6

Agents consuming the SDK can now observe rate-limit utilization and reset times in real time, enabling intelligent work-deferral and priority-based queueing strategies. Sonnet 4.6 as the model tier increases throughput ceiling, making observability more actionable.

2.1.45 Plugin-driven agent capability discovery untested
--add-dir plugins/marketplaces + spinnerTipsOverride

Agents can now bulk-load plugin catalogs and marketplace definitions from directories, enabling dynamic capability composition. spinnerTipsOverride makes long operations more agent-transparent (can be used to signal phase/status in feedback loops).

2.1.47 Background agent fan-out with zero-copy result capture untested
ctrl+f keybinding + last_assistant_message hook field + ESC-detach for background agent lifecycle

The decoupling of background agent termination (ctrl+f to kill all) from main-thread control (ESC still works), combined with direct access to final answers in hook payloads, enables a cleaner multi-agent orchestration pattern where external systems can collect agent results without parsing transcripts.

2.1.47 Live workspace dashboard with agent-synchronized context untested
added_dirs statusline field + last_assistant_message hook field + external script integration

External orchestration scripts can now watch both the active workspace context (via added_dirs in statusline JSON) and the latest agent outcomes (via last_assistant_message), enabling a dashboard that surfaces project context alongside agent-derived insights in real time.

2.1.49 Isolated Agent Orchestration with State Enforcement Untested — requires runtime…
--worktree + isolation: "worktree" + background: true + ConfigChange hook

Worktree-backed subagents running as persistent background tasks can safely parallelize work across branches without merge conflicts or stale-state issues. The ConfigChange hook enables parent orchestrators to react to policy/constraint changes mid-task, allowing enterprise-grade enforcement and task abortion if…

2.1.49 Simple Mode with Direct File Mutation for Embedded/Lightweight Agents Untested — requires validation of…
File edit tool now in simple mode + background: true + plugin settings.json defaults

Lightweight agents deployed in resource-constrained or embedded contexts (e.g., CI/CD pipelines, embedded CLI, SDK integrations) can now edit files directly without Bash, reducing tool surface and permission footprint. Plugins shipping settings.json allows distributing agents with sensible defaults, enabling…

2.1.51 Remote-Controlled Agent Factories in CI/CD untested
claude remote-control + CLAUDE_CODE_PLUGIN_GIT_TIMEOUT_MS + CLAUDE_CODE_ACCOUNT_UUID/CLAUDE_CODE_USER_EMAIL/CLAUDE_CODE_ORGANIZATION_UUID

External build systems (GitHub Actions, GitLab CI, Jenkins) can now spawn ephemeral Claude Code instances with full account context, enabling agents to reason over code generation tasks within CI/CD pipelines without manual API orchestration. The configurable git timeout and account UUID env-vars eliminate cold-start…

2.1.51 Enterprise Deployment with OS-Level Settings Synchronization untested
Macros/Windows Registry managed settings + custom npm registries + BashTool login-shell optimization

Organizations can now centrally manage Claude Code configuration (registry mirrors, timeout policies, workspace trust rules) via macOS plist or Windows Registry, reducing per-machine setup variance. BashTool's automatic login-shell skip further improves command execution performance in enterprise CI/CD contexts where…

2.1.51 Observability-First Agent Telemetry in Distributed Orchestration untested
CLAUDE_CODE_ACCOUNT_UUID + CLAUDE_CODE_USER_EMAIL + CLAUDE_CODE_ORGANIZATION_UUID

Agents spawned in parallel or distributed environments (Kubernetes, serverless) can now tag telemetry with account metadata *before* the first API call, eliminating the race condition where early telemetry events lack account context. This unblocks coherent observability over agent cohorts and audit trails for…

2.1.59 Persistent Agent Context Chains untested
auto-memory (/memory) + per-subcommand "always allow"

Auto-memory captures intermediate learnings (schema, patterns, decision rules) from agents, while smarter permission handling reduces gate friction on compound bash pipelines. Together they enable multi-turn agent sequences to build context momentum without manual state resets or permission fatigue.

2.1.59 Copy-Driven Code Review + Extraction untested
/copy + auto-memory (/memory)

Interactive code block selection (via /copy) paired with auto-memory creates a tighter loop for code review, refactoring, and artifact extraction. Instead of manual copy-paste friction or generic block exports, agents can select precise code regions and persist the selection context for replay or variation.

2.1.63 Code quality gate via /simplify batch untested
/simplify command + /batch command + HTTP hooks

Agents can now chain code review and simplification into repeatable workflows. With /batch, an agent can apply /simplify to multiple files in a single pass and POST the results to an external system (CI/CD, Slack, code review queue) via HTTP hooks, enabling async code-quality gates without custom shell scripts.

2.1.63 Distributed MCP setup with OAuth fallback untested
HTTP hooks + Manual URL paste fallback (OAuth) + Project configs shared across worktrees

Long-running agents in CI/CD or headless environments frequently hit MCP OAuth redirects that fail silently. The new manual URL paste fallback + HTTP hooks enable an agent to recover from failed MCP authentication by prompting a human (or external service) to paste the callback URL, while HTTP hooks let the agent…

2.1.63 Stateful batch jobs with subagent context stripping untested
/batch + Improved memory usage in subagent sessions + HTTP hooks

Large batch jobs spawn multiple subagents, each accumulating context. The new subagent context-stripping optimization prevents unbounded memory growth. HTTP hooks enable status updates to external systems without polling. Together, agents can run large, long-lived batch pipelines (e.g., artifact migration, bulk…

2.1.69 Instrumented Multi-Worktree Orchestrator untested
InstructionsLoaded hook event + agent_id/agent_type hook fields + worktree status-line hook field

These three hook additions together give an orchestrator a complete observability surface for the first time: you can confirm exactly when each agent loaded its instruction context, which agent ID is running in which worktree, and what branch/path that worktree maps to — all in the hook layer before any tool calls…

2.1.69 Clean-Room Agent Git Workflow untested
includeGitInstructions setting + CLAUDE_CODE_DISABLE_GIT_INSTRUCTIONS env var + ${CLAUDE_SKILL_DIR} variable

Suppressing Claude's built-in git instructions removes the default commit/PR prose from the system prompt, freeing budget for a compact, repo-specific git skill that references its own directory via ${CLAUDE_SKILL_DIR}. The combination lets an agent fleet enforce a single canonical git workflow (e.g., Conventional…

2.1.69 Hot-Patch Plugin Dev Loop with git-subdir Sources untested
git-subdir plugin source type + /reload-plugins command

git-subdir lets you point a plugin at a subdirectory of a monorepo without extracting it to its own repo. Combined with /reload-plugins, you get an inner loop where plugin edits in the monorepo are immediately activatable in a running session — no restart, no separate repo management.

2.1.72 Model-differentiated parallel worktree agents untested
ExitWorktree tool + model parameter on Agent tool

The two primitives together make branch-isolated, model-differentiated subagent graphs practical. Before 2.1.72 you could enter a worktree but not cleanly exit it programmatically, and the model param was missing so every subagent used the same model regardless of task weight. Now an orchestrator can spin up a cheap…

2.1.72 SSH-safe artifact extraction pipeline untested
/copy → w key + CLAUDE.md HTML comment hiding

Over SSH, clipboard-dependent extraction is unreliable or unavailable. The w key in /copy writes selections directly to a file, making artifact extraction clipboard-independent. Pairing this with CLAUDE.md HTML comment hiding means operators can embed private routing notes, prompt suppressors, or environment hints…

2.1.72 Inline-scoped plan-mode triage with auto-approval reads untested
/plan <description> optional arg + lsof/pgrep/fd/fdfind auto-approval allowlist additions

/plan <description> lets an orchestrating agent or human pre-seed a plan-mode session with a concrete problem statement and start immediately, skipping the blank-prompt orientation loop. Combined with the six newly auto-approved read-only tools (lsof, pgrep, fd, etc.), the plan-mode agent can do broad environment…

2.1.73 Multi-provider agent fleet with pinned models untested
modelOverrides setting + fixed subagent model downgrades + improved SSL error guidance

Teams deploying agents across heterogeneous LLM providers (corporate Bedrock, Vertex, Foundry, public Claude API) can now pin model versions per provider without silent downgrades or connectivity pain. This enables mixed-provider agent fleets with deterministic behavior.

2.1.73 Reliable multi-session batch processing in monorepos untested
Fixed Bash output loss in multi-session scenarios + fixed skill directory deadlock + fixed permission prompt freezes

Monorepos with large .claude/skills/ directories and concurrent session execution (parallel agent batches, repeated CLI runs in CI) now have stable Bash I/O, skill reloads, and permission-prompt handling. This unblocks automation-heavy agent workflows.

2.1.73 Interrupt+resume with conversation rewind for long-running agents untested
Improved Up arrow restoration + /resume session picker fix

Long-running agents can now be safely interrupted mid-turn (e.g., to fix a typo, reconsider a direction, or poll results) and resumed from the exact conversation state. This reduces the need to discard work or restart from scratch.

2.1.74 Configurable Memory Audit & Optimization Loop untested
/context (enhanced with actionable suggestions) + autoMemoryDirectory + CLAUDE_CODE_SESSIONEND_HOOKS_TIMEOUT_MS

Long-lived agents can now observe memory state, configure where it persists, and control cleanup timing—unlocking automated memory governance for sustained workflows.

2.1.74 Managed Policy Compliance & Model Flexibility untested
Fixed managed policy ask rules + Fixed full model IDs in agent configs

Security policy enforcement and model specification are now consistent across agent frontmatter, JSON config, and CLI—enabling reliable multi-agent teams under managed settings.

2.1.75 Multi-artifact reasoning over large specification sets untested
Opus 4.6 1M context window + Memory last-modified timestamps

The 1M context window removes the working-memory ceiling that previously forced agents to chunk specification ingestion across turns or delegate to sub-agents. Combined with memory freshness tracking, an agent can now load a full specification corpus (dependency graphs, API docs, schema definitions, request/response…

2.1.75 Automated permission auditing with hook source transparency untested
Hook source display in permission prompts + improved async hook messaging

Operators running unattended agents now see exactly which system layer (settings/plugin/skill) is requesting resources. This transparency unlocks automated permission auditing: an agent can log permission patterns, detect unexpected hook sources, and flag configurations that need review without losing context about…

2.1.75 Session continuity for long-running investigation chains untested
Session name display on prompt bar + Memory last-modified timestamps + Fixed /resume for session names

Long investigations (incident response, architecture review, competitor analysis) now have visual continuity. The prompt bar shows the session name, memory freshness is transparent, and /resume preserves session identity across forks. An agent can fork for a risky hypothesis, resume the parent if the fork fails, and…

2.1.76 MCP-gated credential injection untested
MCP elicitation + ElicitationResult hook

MCP servers that need one-time secrets (API keys, OAuth tokens, approval codes) previously had to embed that collection in the tool description or rely on the agent to prompt the user. The elicitation capability lets the MCP server own that dialog, and ElicitationResult lets a hook layer validate or transform the…

2.1.76 Post-compaction state sync for long-running monorepo agents untested
PostCompact hook + worktree.sparsePaths setting

Parallel monorepo agents running in sparse worktrees accumulate long contexts that eventually compact. Before 2.1.76, there was no reliable signal to trigger an external state refresh after compaction — summary files, vector stores, or cross-agent coordination objects could drift silently. PostCompact provides the…

2.1.76 Named sessions for multi-agent orchestration traceability untested
-n / --name <name> flag + PostCompact hook

When an orchestrator launches many parallel subagents, session names in logs and Remote Control dashboards default to opaque identifiers. The --name flag lets the orchestrator stamp each subagent with a semantic name (e.g., the task ID or target file path) at launch. When a subagent compacts, the PostCompact hook can…

2.1.77 Fast Agent Fork-Resume with Higher Capacity untested
/branch (clearer fork semantics) + SendMessage({to: agentId}) auto-resume + --resume 45% faster loading + output token ceiling 128k

The combination of renamed /branch command (semantic clarity), background auto-resume via SendMessage, and dramatically faster session recovery enables agents to fork into parallel analysis streams without losing context or waiting for expensive session reloads. Higher token ceilings mean forked agents can produce…

2.1.77 Sandboxed Read-Access Recovery for Iterative Exploration untested
allowRead setting + Pre-existing denyRead sandbox controls

The new allowRead primitive unlocks selective permission recovery within broad filesystem denials, enabling agents to iteratively explore restricted regions without wholesale permission changes. This reduces security friction in exploration-heavy tasks while maintaining isolation bounds.

2.1.77 Lossy-to-Lossless Response Recovery via /copy N untested
/copy N indexed retrieval + existing multi-turn memory + feedback loop

The new /copy N command enables agents and users to selectively recover earlier assistant responses without re-running turns or searching linearly. Combined with multi-turn memory, this supports workflows where an earlier response proved more useful than a recent refinement, or where parallel response branches need…

2.1.78 API Error Recovery with Graceful Degradation untested
StopFailure hook + line-by-line response streaming

Rate limits and transient API failures currently kill entire workflows silently. By detecting failure via hook and streaming responses line-by-line, agents can offer partial results mid-task and queue retries instead of losing work.

2.1.78 Plugin-Shipped Agents with Persistent Configuration untested
${CLAUDE_PLUGIN_DATA} + effort, maxTurns, disallowedTools frontmatter

Plugins can now ship agents that retain settings across updates and enforce execution constraints without global config pollution. This enables curated, reusable agent templates distributed via plugins.

2.1.78 Real-Time Task Progress with Custom Model Routing untested
line-by-line response streaming + ANTHROPIC_CUSTOM_MODEL_OPTION

Agents can stream output line-by-line while routing model selection via environment variable, enabling dynamic cost/quality trade-offs within a single session and real-time visibility into multi-step work.

2.1.79 Device-agnostic research handoff untested
/remote-control + AI-generated session titles

Extended research tasks can now branch from terminal/VSCode to browser mid-stream, with auto-titled sessions enabling quick context recovery on phone or secondary device without losing chat history.

2.1.79 Plugin seed composition at scale untested
Multi-path CLAUDE_CODE_PLUGIN_SEED_DIR + turn-duration visibility

Multi-directory plugin seeds enable tiered plugin loading (org-wide, team, project, personal) while turn-duration telemetry reveals if plugin composition is slowing startup or turn latency.

2.1.80 Event-Driven Skill Dispatch with Effort Tuning untested
effort frontmatter + --channels (MCP push)

MCP servers can now push events into the session, and skills can declare their computational intensity upfront. This unlocks a reactive agent pattern where external events trigger skills at the effort level that matches their resource budget.

2.1.80 Rate-Limited Skill Orchestration untested
rate_limits visibility + effort frontmatter + skill definition

Agents can now see Claude.ai rate limits (5-hour and 7-day windows) in real time. Skills can declare effort level upfront. Together, these enable a skill orchestrator that checks remaining budget before invoking expensive skills, degrading gracefully to lighter effort levels when nearing limits.

2.1.80 Self-Hosted Plugin Registry via Settings untested
source: 'settings' for plugins + settings persistence

Plugins can now be declared inline in settings.json, removing the need for external marketplace registration. This enables version-controlled, self-hosted plugin catalogs that travel with the repo.

2.1.81 Orchestrated approval routing untested
--bare + --channels

Agents can now route work deterministically: fast, stateless sub-steps run headless (--bare without approval prompts), while sensitive operations requiring human oversight can enable --channels to surface approvals on mobile in real time. This decouples the approval path from the execution path — approval latency no…

2.1.81 Scripted automation without auth machinery untested
--bare

Backend workflows and CI agents have struggled to run Claude Code without OAuth setup, credential rotation, or interactive auth. --bare removes all auth machinery (oauth, keychain, interactive key input) and requires only ANTHROPIC_API_KEY in the environment — the minimal subset for stateless, deterministic automation.

2.1.83 Reactive Monorepo Environment Switcher untested
CwdChanged hook event + FileChanged hook event + CLAUDE_CODE_SUBPROCESS_ENV_SCRUB=1

CwdChanged fires when the agent navigates between packages in a monorepo; combined with FileChanged watching .envrc or package.json, a hook can automatically run direnv, reload tool configs, and scrub stale credentials from subprocess envs — making multi-package agentic sessions self-configuring rather than requiring…

2.1.83 Self-Bootstrapping Agent with Credential-Safe Subprocess Isolation untested
initialPrompt frontmatter + CLAUDE_CODE_SUBPROCESS_ENV_SCRUB=1 + sandbox.failIfUnavailable

An agent that declares its own initialPrompt is fully self-contained — no wrapper script needed to seed it. Pairing that with CLAUDE_CODE_SUBPROCESS_ENV_SCRUB=1 ensures the agent's Bash tool and MCP stdio servers never inherit ambient credentials from the launch shell. Adding sandbox.failIfUnavailable: true prevents…

2.1.83 Multi-Team Policy Layering with Per-Team Credential Isolation untested
managed-settings.d/ drop-in directory + Plugin manifest.userConfig / sensitive: true

managed-settings.d/ lets a platform team ship a baseline policy fragment while individual product teams overlay their own fragments without editing shared files. Plugin manifest.userConfig with sensitive: true lets each team's plugins prompt for their own secrets at enable time and store them in the OS keychain — so…

2.1.84 Task-Gated Orchestration with Audit Log untested
TaskCreated hook + WorktreeCreate HTTP hook (hookSpecificOutput.worktreePath)

TaskCreated fires at task creation and WorktreeCreate HTTP now returns the worktree path to the hook server — together they let a remote hook server intercept both ends of an agentic sub-session: the task that owns the work, and the worktree it runs in. This closes the loop on remote orchestration: you can correlate…

2.1.84 Third-Party Model Capability Override for Bedrock/Vertex Pipelines untested
ANTHROPIC_DEFAULT_{OPUS,SONNET,HAIKU}_MODEL_SUPPORTS env-var + CLAUDE_STREAM_IDLE_TIMEOUT_MS env-var

When running Claude Code against Bedrock or Vertex with pinned model ARNs, capability detection and streaming reliability are the two most common breakage points in automated pipelines. These two env-vars address both simultaneously without code changes: MODEL_SUPPORTS tells the harness which effort/thinking…

2.1.85 Autonomous Permission-Satisfied Tool Chains untested
PreToolUse updatedInput + Hook if field conditional filtering

Agents can now route tool execution through pre-answered questions without interactive prompts, and filter noisy permission checks via conditional hook rules—enabling fully autonomous multi-tool workflows in headless/scheduled contexts.

2.1.85 Deep-Link Prompt Expansion for Skill Context Injection untested
Deep link query expansion (5k chars) + Transcript timestamp markers + Hook if field

Agents can now embed up to 5,000 characters of skill context, directives, and prior work state into deep links, and scheduled routines will emit timestamps to anchor the injected context in transcript history.

2.1.85 Headless MCP Multi-Server Routing untested
CLAUDE_CODE_MCP_SERVER_NAME + CLAUDE_CODE_MCP_SERVER_URL env-vars + PreToolUse updatedInput

A single MCP headersHelper script can now route multiple servers by name/URL env-vars, and PreToolUse hooks can satisfy MCP auth questions programmatically—enabling zero-touch delegation to specialized servers in cloud/headless environments.

2.1.86 Multi-Instance Session Aggregation + Proxy Intelligence untested
X-Claude-Code-Session-Id header + Read tool token optimization

The session-ID header + reduced token overhead on file mentions enable smarter proxy-layer session tracking without body parsing, allowing orchestration systems to correlate work across multiple Claude Code instances while staying within token budgets.

2.1.86 Robust Long-Session Agent Feedback Loop untested
Fixed --resume + out-of-memory crash fix + /feedback reliability

2.1.86 hardens the session resume and feedback mechanisms, allowing agents to safely capture, persist, and replay multi-hour runs with /feedback without crashing or losing state on recovery.

2.1.89 Interactive Permission Gates for Headless Agents untested
"defer" permission decision (PreToolUse hook) + -p --resume + PermissionDenied hook + named subagents in @ mention

Long-running orchestration workflows can now pause at permission boundaries, notify human operators, and resume with fresh evaluation — without blocking or losing session state. Named subagents in mention discovery let agents route permission decisions to specialized auditors.

2.1.89 Flicker-Free Long-Running Orchestration Loops untested
CLAUDE_CODE_NO_FLICKER=1 + MCP_CONNECTION_NONBLOCKING=true + improved bash tool warnings

Scheduled cloud agents and persistent orchestration loops can now render cleanly without terminal flicker and skip MCP connection blocking, reducing visual noise and latency in observability dashboards and long-running monitor loops.

2.1.90 Offline-resilient marketplace + guided onboarding untested
CLAUDE_CODE_PLUGIN_KEEP_MARKETPLACE_ON_FAILURE + /powerup

Remote teams and air-gapped environments can now learn Claude Code interactively without disrupting marketplace availability when network fails. Combines survivable offline caching with discoverable feature training.

2.1.91 Large-Schema Discovery Loop untested
_meta["anthropic/maxResultSizeChars"] + Plugin bin/ executables

Agent-driven database or API schema ingestion typically fails on truncated output. By persisting full schemas (up to 500K) alongside lightweight custom tools (plugin executables), agents can now perform lossless schema import, cross-database comparison, and generation of integration glue code without round-tripping to…

2.1.91 Sandboxed Skill Execution in CRC Graph Adjudication untested
disableSkillShellExecution + multi-line deep links

CRC v3 adjudication workflows often call out to skills to fetch or verify facts. The ability to disable inline shell execution in skills (security boundary) + multi-line deep links (structured prompts to agents) enables safer, richer agent choreography where untrusted or third-party skills can be executed under…

2.1.92 Policy-Gated Multi-Model Orchestration untested
forceRemoteSettingsRefresh + Bedrock setup wizard + /cost per-model breakdown

Remote managed settings enforce policy compliance on startup; Bedrock wizard enables seamless 3rd-party model onboarding; enhanced /cost visibility allows agents to route workloads by model economics and compliance posture in real time.

2.1.92 Session Reliability in Long-Running Agent Batches untested
Subagent spawn fix + tmux window resilience + prompt cache expiry hint

The "Could not determine pane count" failure was blocking multi-hour batch jobs when tmux windows died; the fix restores spawn reliability. Cache expiry hint helps agents estimate token cost before next turn, enabling graceful degradation.

2.1.92 Cost-Aware Session Management with Interactive Introspection untested
/cost per-model breakdown + /release-notes version picker + improved Write diff speed

Agents can now see granular cost-per-model and cache-hit rates, and the version picker lets them inspect how changes in Claude Code releases affected their performance. Faster Write diff computation reduces per-artifact overhead.

2.1.94 Multi-Region Bedrock Inference + Session Instrumentation untested
CLAUDE_CODE_USE_MANTLE=1 + hookSpecificOutput.sessionTitle

Agents running in regulated or cost-sensitive environments (healthcare, finance, on-premise) need to route inference to customer-owned Bedrock endpoints while maintaining audit trails. Dynamic session titling from hooks unlocks traced, multi-tenant reasoning pipelines where each invocation is named after its…

2.1.94 Stable Plugin Skill Composition + Hook-Driven Output Control untested
Plugin skill frontmatter name precedence + keep-coding-instructions + hookSpecificOutput.sessionTitle

Plugin skills installed via "skills": ["./"] now have stable invocation names across package managers, and can opt into preserving coding context (instructions) in rendered output. Combined with dynamic session titling, agents can compose reusable skill stacks with deterministic names, audit-friendly output, and…

2.1.94 Silent Failure Elimination in Long-Agentic Sessions untested
429 error surfacing + scrollback fix + CJK stream-json fix + --resume cross-worktree + plugin hook visibility

Long-running agent workflows (multi-day data pipelines, batch processing) previously silenced rate-limit errors, corrupted multibyte streams, and lost scroll context. 2.1.94 surfaces these issues immediately and makes session resumption from sibling worktrees seamless, eliminating invisible failure modes that trap…

2.1.97 Compact Session Monitoring for Long Agent Runs untested
Ctrl+O focus view toggle + refreshInterval status line setting + ● N running indicator

Multi-stage orchestrations (parent spawning subagents, chained delegations) need lightweight session inspection without burning context. The focus view collapses prompt/tool/response into one viewport; refreshable status line enables polling without re-reading; live agent count indicator surfaces orchestration depth…

2.1.97 Worktree-Isolated Subagent Orchestration with Transparency untested
workspace.git_worktree status line field + subagent worktree isolation fix + /agents live indicator

Sparse worktrees isolate subagent filesystem mutations but previously leaked working directory back to parent, risking silent contamination. The 2.1.97 fix + workspace.git_worktree JSON field enable parent to verify isolation state in real time and route work accordingly.

2.1.97 Cedar Policy Authoring with Live Syntax Feedback untested
Cedar syntax highlighting + CJK sentence punctuation completion

Cedar (AWS IAM policy language) is domain-specific; full highlighting is new in 2.1.97. CJK punctuation support removes friction for multi-language policy teams. Together, they enable fluent authoring without space-padding syntax completion.

2.1.98 Reactive Build-Monitor Agent untested
Monitor tool + workspace.git_worktree status field

Monitor gives agents a native event stream from background processes; workspace.git_worktree lets a hook or status-line integration know which worktree the agent is operating in. Together, a multi-worktree agent setup can launch a background build/test process in each worktree and route Monitor events back to the…

2.1.98 Sandboxed Script-Capped Agentic CI untested
CLAUDE_CODE_SUBPROCESS_ENV_SCRUB (PID namespace isolation) + CLAUDE_CODE_SCRIPT_CAPS + Monitor tool

Subprocess sandboxing with PID namespace isolation prevents runaway child processes from escaping their session boundary; CLAUDE_CODE_SCRIPT_CAPS provides a hard cap on total script invocations so a misbehaving agent loop cannot exhaust the environment. Pairing both with Monitor lets an operator-controlled harness…

2.1.101 Automated Team Onboarding with Remote Agents untested
/team-onboarding command + /ultraplan auto-environment + OTEL tracing env-vars

Team ramp-up guides paired with auto-provisioned remote environments enable agents to train peers asynchronously without manual setup. Tracing visibility lets coordinators audit what knowledge was transferred and where agents struggled.

2.1.101 Enterprise Proxy with Traceable Cloud Orchestration untested
CLAUDE_CODE_CERT_STORE=bundled env-var + /ultraplan auto-environment + improved refusal/rate-limit error messages

Teams behind enterprise TLS proxies can now trust the OS CA store (or pin to bundled), auto-create remote sessions without web friction, and see transparent error messages when hitting rate limits or API refusals — enabling agents to retry or escalate intelligently.

2.1.101 Resilient Multi-Agent Session Recovery untested
Hardened --resume/--continue chain recovery + isolated worktree file access fixes + permissions.deny override fixes

Subagents can now safely resume interrupted sessions from deep isolation, read/write files in their own worktree without permission barriers, and have deny rules properly enforced. This unlocks durable conversation recovery across agent handoffs without loss or privilege escalation.

2.1.105 Persistent multi-branch session fusion untested
EnterWorktree path parameter + PreCompact hook decision:block

Agents managing multiple concurrent PR branches can now navigate existing worktrees by path (not re-entry) and guard compaction when critical session state spans branches, enabling continuous session semantics across branch switches without state loss.

2.1.105 Plugin observability layer untested
Plugin monitors manifest key + improved MCP large-output truncation recipes

Plugins can now auto-arm background monitors at skill invocation, paired with format-aware truncation prompts (jq, chunk-size recipes), enabling plugins to maintain live observability without manual hook plumbing or UI state loss.

2.1.108 Latency-Adaptive Polling Agent untested
FORCE_PROMPT_CACHING_5M + ENABLE_PROMPT_CACHING_1H + /recap

Agents can now trade cache freshness for latency by selecting cache TTL per task phase. 5-minute TTL suits rapid-feedback loops (deployment polls, test retries); 1-hour TTL suits stable-state tasks (report generation, batch processing).

2.1.108 Self-Discovering Agent Tool Chains untested
Model slash-command discovery + /recap + improved /resume defaults

Agents spawn with built-in slash commands automatically available for discovery (no teach-in required), and session recovery is seamless by default (local-project focus, recap context). This reduces setup friction when agents fork, resume, or run across repo boundaries.

2.1.108 Cache-Aware Error Recovery untested
Improved error messages (rate limit vs. plan limits) + FORCE_PROMPT_CACHING_5M + startup cache warning

Agents can now distinguish transient limits (rate limiting, recoverable with backoff and tighter cache) from terminal limits (plan exhaustion). Startup warnings surface cache misconfiguration early, preventing silent cache bypass.

2.1.110 Async-aware scheduled agent with mobile notifications untested
--resume/--continue task resurrection + push notification tool + Remote Control slash-command support

Scheduled agents (via --resume) can now resurrect unexpired tasks, send mobile push notifications when decisions require user input, and execute remote commands from mobile/web clients. This unlocks fully async agent delegation where the agent runs on a schedule, makes progress, and notifies the user to review/approve…

2.1.110 Flicker-free interactive refactoring with scheduled resume checkpoints untested
/tui fullscreen + /focus command + --resume/--continue task resurrection

Agents can now work in a flicker-free TUI mode (/tui fullscreen), toggle focus to hide distracting elements (/focus), and persist checkpoints by exiting and resuming via --resume. This unlocks long-running interactive refactoring sessions where the agent can take clean breakpoints without losing context.

2.1.110 Distributed-trace-aware cloud agent observability pipeline untested
TRACEPARENT/TRACESTATE env support + SDK/headless sessions

Cloud-orchestrated agents can now link their execution traces to parent traces via environment variables, enabling end-to-end observability when agents are spawned by a parent service or orchestrator (e.g., a Temporal workflow, a Dagster job, or a cloud function).

2.1.113 Restricted Remote Agent Orchestration untested
sandbox.network.deniedDomains + /extra-usage on Remote Control + Bash wrapper-matching security rules

Multi-tier agent sandboxing now possible on remote (mobile/web) clients — a desktop session can spawn subagents running against filtered network surfaces, with explicit /extra-usage provisioning visible to remote operators, and deny rules that catch sophistication attempts (wrapped sudo/env escapes).

2.1.113 Full-Screen Responsive Selection Prompting untested
Fullscreen Shift+↑/↓ selection scrolling + Ctrl+A/Ctrl+E readline behavior + slash/@ menu flush positioning

Multiline prompt composition in fullscreen mode now responds to readline keybindings and viewport-aware selection, matching muscle memory from terminal shells and closing the UX gap for agents drafting complex prompts without intermediate escapes to the terminal.

2.1.118 Agent-driven MCP verification in auto-mode untested
type: "mcp_tool" hooks + "$defaults" auto-mode token

Hooks invoking MCP tools directly, combined with the ability to layer custom auto-allow rules alongside built-ins, unlock verifier hooks that can consult external services (MCP-backed databases, APIs, permission servers) before accepting or auto-allowing a user's command.

2.1.118 Plugin-branded agent environments with release tagging untested
custom themes via plugin themes/ directory + claude plugin tag command

Plugins can now ship custom themes, and developers have a version-validated release mechanism (claude plugin tag) to ensure theme definitions are stable across plugin versions. This enables agent scaffolding tools to brand their plugin environments (color palette, UI theme) and release them alongside code updates.

2.1.119 Observability-Driven Tool Chain Tuning untested
PostToolUse/PostToolUseFailure duration_ms + CLAUDE_CODE_HIDE_CWD + custom prUrlTemplate

Hook-level latency telemetry enables agents to profile and optimize tool chains in secure multi-tenant or branded environments without exposing working directories.

2.1.119 Enterprise SCM + Persistent Config for Repeatable Agents untested
--from-pr (extended to GitLab/Bitbucket/GHE) + /config settings persistence + --agent <name> permissionMode

Multi-SCM PR ingestion combined with persistent agent configuration enables repeatable, org-standardized code review and CI workflows across diverse git platforms without manual CLI flag management.

2.1.119 PowerShell Automation with Hook-Level Performance Feedback untested
PowerShell auto-approval + PostToolUse duration_ms + parallel MCP server reconfiguration

PowerShell auto-approval removes manual permission gates in Windows-native environments; hook observability and parallel MCP startup accelerate complex multi-tool automation pipelines.

2.1.120 Review Gate in Agent-Driven CI/CD untested
claude ultrareview [target] + ${CLAUDE_EFFORT} + AI_AGENT

Agents can now run structured code review as a pipeline gate, scaling review effort with the task's urgency level and attributing all review traffic to the agent for audit/cost tracking.

2.1.120 Effort-Scaled Skill Provisioning untested
${CLAUDE_EFFORT} + PowerShell shell fallback (Windows)

Skills can now adapt their behavior to the parent agent's effort level, enabling same-skill deployment across quick-turnaround and deep-analysis use cases. The PowerShell fallback removes platform-specific skill branching logic.

2.1.121 Non-interactive Subagent CI/CD Orchestration Untested
CLAUDE_CODE_FORK_SUBAGENT=1 (non-interactive) + alwaysLoad (MCP config) + --dangerously-skip-permissions

Together these enable fully automated subagent spawning in CI/CD pipelines without blocking on permission prompts or deferred tool resolution, allowing agents to coordinate across build steps with eager tool availability.

2.1.121 Terminal-Driven Agent Configuration + Hotkey Execution Untested
/skills search box + CLAUDE_CODE_FORK_SUBAGENT=1 + hookSpecificOutput.updatedToolOutput

Agent developers can now discover and configure skills interactively in the terminal, then fork subagents to execute them while post-processing tool results through hooks—collapsing the edit-test-deploy cycle.

2.1.121 MCP OAuth + Observability Audit Trail for Agent Ops Untested
mcp_authenticate with redirectUri + OTEL_LOG_USER_PROMPTS (OpenTelemetry) + NO_PROXY managed-settings fix

Agents can now authenticate to MCP services (e.g., GitHub, Slack) via custom redirect schemes and emit full audit traces (system prompt, token counts, finish reasons) to observability backends—essential for multi-tenant agent platforms with compliance audit requirements.

2.1.122 PR-indexed session recovery + Bedrock cost-tier tuning untested
/resume PR URL search + ANTHROPIC_BEDROCK_SERVICE_TIER env var

Multi-PR review sessions waste time rediscovering which Claude Code session created which artifact. Combined with Bedrock tier selection, agents and users can now (1) jump to a PR's originating session to inspect build logs/reasoning without re-reading chat, and (2) downgrade to flex tier mid-review to reduce cost on…

2.1.122 At-mention logging + session branching for audit trails untested
OpenTelemetry claude_code.at_mention log event + /branch stability fix

Agents that use @-mentions to invoke tools now have structured, observable mention resolution logged. The /branch fix (fork corruption from rewound timelines) ensures that audit trails from complex branching workflows no longer corrupt the tool-use chain. Together, they enable tracing which tool was invoked, when, and…

2.1.129 Exotic Terminal + Skill-Gated Model Inference untested
CLAUDE_CODE_FORCE_SYNC_OUTPUT=1 + skillOverrides setting

Agents running inside constrained CI/automation environments (Emacs eat, Docker, GitHub Actions on ARM) often fail silently due to terminal output buffering and unintended skill inference. Combining forced synchronous output with skill availability gates enables reliable, auditable automation.

2.1.129 Model Discovery Control for Multi-Tenant / Air-Gapped Deployments untested
CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY=1 (opt-in) + deniedMcpServers pattern fixes

Third-party deployments (Vertex, Bedrock, Foundry, custom gateways) now explicitly control whether the model picker discovers models via gateway /v1/models. The fix to deniedMcpServers wildcard matching (case-insensitive hostnames) enables precise network boundary enforcement.

2.1.129 Skill Visibility Governance for Collaborative / Educational Agents untested
skillOverrides setting (modes: off, user-invocable-only, name-only)

Not a multi-primitive workflow, but a single newly-working primitive that enables collaborative and educational agents to enforce skill governance. The skillOverrides modes allow fine-grained control: off hides skills completely, user-invocable-only prevents model inference but allows human / calls, and name-only…

2.1.132 Session-Correlated Audit Log with Native Scrollback Persistence untested
CLAUDE_CODE_SESSION_ID + CLAUDE_CODE_DISABLE_ALTERNATE_SCREEN=1

Agents running in headless, containerized, or remote environments need durable output trails that survive terminal reconnects and can be audited post-session. Disabling alternate-screen keeps all output in native scrollback; stamping each bash subprocess's output with CLAUDE_CODE_SESSION_ID enables post-hoc log…

2.1.133 Draft-Preserving Agent Isolation untested
worktree.baseRef + effort.level (JSON/env)

Multi-agent workflows often need a parent agent to hand off draft code (uncommitted local changes) to a child agent in an isolated worktree. Previously, isolation always branched from remote origin, forcing commits or loss of work. Now agents can set worktree.baseRef: "head" to preserve local HEAD, and child agents…

2.1.133 Effort-Aware Hook Automation untested
$CLAUDE_EFFORT environment variable + Hooks receiving effort.level JSON input

Hooks are the primary escape hatch for build systems, formatters, and pre-commit checks. Previously, hooks had no way to know the user's intent (quick smoke test vs. deep audit). Now hooks can branch on $CLAUDE_EFFORT or parse the JSON input to scale validation: skip expensive checks on low effort, run all checks on…

2.1.133 Enterprise Admin Settings Bridge (SDK + CLI) untested
parentSettingsBehavior admin-tier setting + sandbox.bwrapPath / sandbox.socatPath

Organizations running Claude via SDK (managed parent tier) and CLI (user tier) have conflicting sandbox and security policies. Previously, admin-managed settings always won-only (first-wins), forcing CLI users to match parent sandbox paths. Now admins can opt into merge behavior to let CLI users override sandbox…

2.1.136 Enterprise Compliance Audit Loop untested
CLAUDE_CODE_ENABLE_FEEDBACK_SURVEY_FOR_OTEL + settings.autoMode.hard_deny

Sensitive deployments (government, regulated industries) can now wire session quality feedback directly into compliance telemetry pipelines while enforcing immutable permission boundaries that satisfy air-gapped security reviews.

2.1.136 MCP Server Resilience in Multi-Tenant Agent Fleets untested
MCP OAuth refresh token fix + MCP server persistence fix (from hardening)

Multi-server remote MCP fleets previously required daily re-auth and lost server configurations on session clear. These fixes unlock sustained unattended agent operation without stale-credential loops or silent server dropout.

2.1.141 Daemonized Agent Orchestration with Terminal-Independent Signaling untested
terminalSequence hook field + Background agent permission preservation + claude agents --cwd <path>

Backgrounded agents can now emit desktop notifications independently of terminal attachment, preserve their permission context across sessions, and be scoped to a directory. This unlocks orchestration of parallel agent swarms that coordinate via notifications, maintain isolated permission domains, and organize work by…

2.1.141 Workspace-Isolated Multi-Tenant Agent Routing untested
ANTHROPIC_WORKSPACE_ID + CLAUDE_CODE_PLUGIN_PREFER_HTTPS + Background agent permission preservation

Workload identity federation now allows token scoping to a specific workspace, plugin bootstrap works in restricted networks (no SSH), and spawned agents inherit the parent's permission policy. Together these enable multi-tenant agent deployments where each tenant's agents are token-scoped, can bootstrap in zero-trust…

2.1.141 Context-Optimized Multi-Session Issue Tracking untested
/feedback recent sessions + Rewind "Summarize up to here" + claude agents --cwd <path>

Feedback can now include recent sessions from across multiple agent runs, rewind can compress early context while preserving recent turns, and session scoping by directory lets agents focus on task-relevant sessions. Together these enable efficient issue diagnosis across distributed agent work without context…

2.1.142 Isolated Model Experiments with Tailored Worker Sessions untested
--model flag + --effort flag + --dangerously-skip-permissions flag + Opus 4.7 default upgrade

Agents can now spawn isolated background workers pinned to specific models and reasoning effort levels, enabling parallel hypothesis testing (e.g., Haiku fast-explore vs. Opus deep-think) without blocking the main session or requiring user permission re-prompting.

2.1.142 Persistent Graph-Synced Agents with Preconfigured Boundaries untested
--add-dir flag + --settings flag + --permission-mode flag + root-level SKILL.md auto-surfacing

Agents can dispatch workers with exact directory scope, pre-loaded settings, and permission lockdown, enabling reliable graph-synced agent swarms where each worker has a bounded execution context and repeatable configuration state.

2.1.142 MCP Tool Farm with Stable Daemon Lifecycle untested
--mcp-config flag + MCP_TOOL_TIMEOUT hardening + daemon sleep/wake resilience

Agents can now provision background workers with explicit MCP server configuration and rely on stable daemon behavior across system sleep/wake cycles, enabling 24-hour agent farms with long-running tool integrations (e.g., database pollers, API monitors) that previously crashed on macOS.

2.1.143 Plugin-safe fleet management in monorepo untested
claude agents --add-dir + claude agents --mcp-config + claude agents --permission-mode + claude plugin disable dependency enforcement

Agents dispatched from the dashboard can now carry uniform MCP server, permissions, and directory context without manual setup, plus plugin dependency enforcement prevents silent breakage when multiple agents share plugin dependencies. This unlocks safe, scalable multi-agent setups in large codebases.

2.1.143 Persistent background workers in worktree-hostile environments untested
worktree.bgIsolation: "none" + /bg preserving --mcp-config, --settings, --plugin-dir

Repos with complex layouts or shared filesystem constraints (monoliths, git-filter-branch artifacts, NFS mounts) can now run background agents directly on the working copy without the overhead of git worktrees, while still preserving full transitive configuration across sleep/wake cycles.

2.1.143 Cost-aware plugin marketplace browsing for context-sensitive agent design untested
/plugin projected context cost + claude agents --model + --effort defaults

Before assembling a fleet of agents, browse the plugin marketplace with cost projections to estimate context footprint, then dispatch workers with matching model/effort tiers. Prevents context bloat and mismatches in heterogeneous agent roles.

2.1.144 Async Research + Resume untested
/resume + elapsed duration display + /model session-only isolation

Long-running research tasks (deep searches, data synthesis, report generation) can now be spawned in background sessions with explicit model choice, detached, and resumed later with full context—returning agents see both the result and the wall-clock time spent, enabling resource-aware orchestration.

2.1.144 Multi-Model Cascade with Session Binding untested
/model session-only + /resume + background isolation

Route tasks to model-specific sessions without global state pollution. Haiku sessions handle intake/routing, Opus sessions handle deep reasoning, and Claude 3.5 Sonnet sessions handle code generation—all coexist, all resumable, all keeping their own model choice locked.

2.1.144 Distributed Telemetry via Duration Badges untested
Elapsed duration display + /resume + background completion notifications

Agent orchestrators can now inspect wall-clock time alongside outputs, enabling cost-to-outcome and resource-allocation reasoning. Background completion notifications (e.g., "Agent completed · 3h 2m 5s") become telemetry signals for routing decisions.

2.1.145 Distributed Agent Session Resurrection & Status Monitoring untested
claude agents --json + agent_id and parent_agent_id OTEL span attributes + claude agents awaiting-input indicator

A subagent orchestrator can now introspect all live Claude sessions, disambiguate parent-child agent relationships through trace attributes, and monitor which agents are waiting for input—enabling tmux-resurrect-style session persistence, automated session pruning, and responsive multi-agent coordination without…

2.1.145 Hook-Aware Task and Cron Lifecycle Management untested
background_tasks in Stop/SubagentStop hook input + session_crons in Stop/SubagentStop hook input + fixed skill context: fork infinite loop

Stop/SubagentStop hooks can now inspect what background tasks and crons a session owns before teardown, enabling hooks to implement graceful cleanup patterns (drain task queues, migrate crons to persistent storage, or abort stale work). The fixed context: fork loop prevents unintended skill re-invocations that would…

2.1.145 Plugin Intelligence Before Installation untested
/plugin pre-installation metadata display + mouse hover and click in suggestion list

A user or automation can now inspect what capabilities a plugin exports (commands, agents, skills, hooks, MCP/LSP servers) before committing to installation, and use fullscreen suggestion navigation to browse plugin catalogs interactively—reducing discovery friction and installation regret.

2.1.152 Autonomous Code Hardening Loop untested
/code-review --fix + disallowed-tools in frontmatter + MessageDisplay hook

Code review suggestions are often ignored or require manual application. By combining automatic fix application, tool-suppression during review-heavy phases, and hook-driven output transformation, agents can run self-contained hardening cycles that review their own work, apply fixes, suppress noise, and surface only…

2.1.152 Dynamic Skill Injection with Session Context untested
SessionStart hook reloadSkills: true + hookSpecificOutput.sessionTitle + /reload-skills

Most skills are static; they can't adapt to the session's purpose or load new capabilities on demand. By having hooks inject skills mid-session and set the session title to reflect the active skill set, agents can dynamically reconfigure their tooling and make their mode transparent to the user.

2.1.152 Suppressed Output Audit Agent untested
MessageDisplay hook + --scope user|project|local marketplace control + /usage large file scanning

Audit and diagnostic agents often produce verbose intermediate logs and intermediate findings that obscure the final result. A MessageDisplay hook can filter or hide diagnostic output while preserving the final report. Combined with scoped marketplace policies and transparent session usage tracking, this enables…

2.1.153 Terminal-Aware Status Dashboards for Background Agents untested
Status line COLUMNS/LINES env vars + /model persistence

Agents running status-line dashboards can now adapt layout and content density to terminal dimensions, and model selection stays stable across agent respawns, enabling rich but responsive CLI monitoring.

2.1.153 Faster Project Onboarding with Smart Dispatch untested
skipLfs option for git sources + claude agents autocomplete with native slash commands + PR column formatting

New agents onboarding large LFS-heavy projects can now skip expensive LFS pulls, receive better skill/command hints, and report PR status more clearly in the dispatch UI.

2.1.157 Repo-Local Skill Dispatch with Context Routing untested
Local plugin auto-load from .claude/skills + agent field honored in settings.json + EnterWorktree mid-session switching

Decoupling skill distribution from the marketplace and binding agent identity to the repo allows polyrepo monoliths and multi-team foundations to inject role-specific toolchains at checkout time, then pivot between specialized agents mid-session without losing conversation state or requiring CLI reauthorization.

2.1.157 Deterministic Multi-Agent Choreography with Filtered Keyword Triggers untested
"Workflow keyword trigger" setting + agent field routing + telemetry tool_parameters logging

Agents orchestrating other agents can disable accidental workflow prompts (removing the "workflow" keyword auto-trigger) while logging deterministic tool invocations, enabling parent orchestrators to reason about and replay agent work without interference from user-typed keywords.

2.1.157 Workstation State Cleanup and Orphan Recovery untested
Worktrees left unlocked after agent finish + fixed background worktree orphaning in 30-day sweep + --resume background subagent reporting

Claude now leaves worktrees in a cleanable state and correctly reports which background agents were running before a crash or reconnect, enabling cleanup scripts and session recovery flows to operate deterministically without stale locks.

2.1.163 Version-gated multi-agent orchestration untested
requiredMinimumVersion / requiredMaximumVersion + CLAUDE_CODE_SESSION_ID for stdio MCP servers

Managed version constraints prevent agent fleet incompatibility while session-ID propagation ensures coordinated MCP state across delegated agents, enabling safe version-locked orchestration in production.

2.1.163 Hook-driven feedback loops in delegated stops untested
hookSpecificOutput.additionalContext for Stop/SubagentStop + hook if: pattern fixes

Stop hooks can now return feedback to Claude without aborting the turn, enabling graceful downstream coordination when a delegated subagent or background task finishes—turn keeps going, hook provides context, no false errors.

2.1.163 Clipboard-enhanced interactive agent annotation untested
"c to copy" shortcut on /btw + /plugin list visibility

Agents can now invoke /btw to cite sources and annotations, then copy them verbatim to clipboard for manual review or downstream tool integration, reducing friction in human-in-the-loop and hybrid-annotation workflows.

2.1.166 Graceful Model Degradation for Agent Orchestration untested
fallbackModel setting + --fallback-model flag + API error retry on fallback model

Agent orchestrators can now seamlessly shift work to secondary models when the primary saturates. Combined with automatic retry-on-fallback, this means background agents no longer block or fail on load; instead they downgrade gracefully and complete reliably.

2.1.166 Token-Budget Tuning for Thinking Models via Environment Control untested
MAX_THINKING_TOKENS=0 env-var + per-model thinking toggle + --thinking disabled flag

Agents can now trade thinking cost for speed by explicitly disabling thinking on default-thinking models (e.g., 3.7+), reclaiming tokens for actual tool invocation. Per-model toggles let agents mix thinkers (reasoning-heavy tasks) with non-thinkers (fast execution tasks) in the same session.

2.1.166 Sandboxed Multi-Session Relay with Messaging Hardening untested
Cross-session messaging hardening (SendMessage no longer carries user authority)

Multi-agent systems that relay requests across sessions can now safely separate trust boundaries. A coordinator session cannot accidentally grant a worker session its permissions; each session retains isolated authority.