v2.1.240

2.1.240

(☝️ Click to see full diff.)

-1,911 tokens

• NEW: Tool Description: Agent (simple usage notes) — Adds concise Agent-tool guidance covering when to delegate, fork behavior, resuming agents, worktree isolation, background execution, parallel launches, and context restrictions.
• Agent Prompt: Security monitor for autonomous agent actions (second part) — Expands Self-Modification protection to an explicit set of agent configuration paths, while treating .claude/worktrees/<name>/ and unlisted project-specific .claude/ directories as ordinary project files unless they contain protected configuration.
• Agent Prompt: Worker fork — Clarifies that “default to forking” guidance belongs to the parent agent and that the fork must execute its directive directly without spawning another agent.
• Tool Description: Snooze (delay and reason guidance) — Prohibits short-interval polling for harness-tracked background work, recommends a 1200-second-or-longer fallback heartbeat, and reserves short cache-preserving delays for external state such as CI runs, deployments, and remote queues.
• Tool Description: Write (read existing file first) — Clarifies that Write is for creating files or fully replacing previously read files, while partial modifications should use Edit.

v2.1.239

2.1.239

(☝️ Click to see full diff.)

+960 tokens

• NEW: Data: SDK MCP server errors field — Documents skipped --mcp-config entries on SDK init frames, including stable error categories, omitted affected servers, Remote Control bridge failures, and CI handling.
• NEW: Data: SendMessage ambiguous recipient display — Restores user-facing “not sent” explanations for inexact or duplicate agent names, unavailable or truncated session searches, and recipients that require exact-name or pinned-identity confirmation.
• NEW: Tool Description: Artifact content host network block guidance — Explains the Artifact content-host allowlist entry for personal and shared cloud environments and forbids republishing until Claude can read the live version.
• NEW: Tool Description: Artifact files guidance — Adds list_files and read_file actions for multi-file Artifacts, defaulting reads to the scratchpad while reserving out_dir for user-requested destinations.
• NEW: Tool Description: Artifact identical resubmission refusal, Tool Parameter: Artifact URL guidance, and Tool Parameter: Artifact force overwrite guidance — Require existing owned Artifact URLs for in-place updates, reject unchanged retries after stale-version conflicts, and reserve forced overwrites for explicit approval to discard a specific newer version.
• REMOVED: Skill: Artifact slides and Skill: Artifact spreadsheet — Remove the standalone editable slide-deck and spreadsheet Artifact skills and their template-preservation workflows.
• Agent Prompt: Security monitor for autonomous agent actions — Adds an unverifiable-deletion-scope soft block for runtime-computed destructive writes against shared or remote state, requiring a transcript-visible resolved target list or literal verified names within the user-approved scope.
• Skill: /insights report output — Reworks the restored /insights flow into an exact unwrapped shareable-report handoff, replaces the additional-context injection with report-header context, and forbids adding, omitting, or rewording any line.
• Tool Description: Artifact and Skill: Artifact design — Make HTML the default Artifact format, allowing Markdown only when a loaded skill explicitly requires it and converting shared Markdown documents into designed HTML pages rather than one-to-one transcriptions.
• Tool Description: Artifact publishing and update guidance; Skill: Artifact PR review, Skill: Design, and Skill: Whiteboard — Move Artifact reads from WebFetch to action: "read", returning owned pages as raw HTML and shared pages as isolated summaries—except same-session Slack-channel pages, which return full untrusted content—while updating dependent workflows to consume saved large-page results safely.
• Tool Description: Artifact publishing and update guidance, Tool Description: Artifact runtime capabilities guidance, and Tool Parameter: Artifact watch actions guidance — Distinguish unavailable, durable-wake, and live watch modes; expand capability examples to live or connected data, shared state, viewer identity, Claude questions, added files, and self-saving pages; and require re-read, merge, and republish handling when local source falls behind a page-authored version.
• Tool Description: Artifact live room guidance — Requires explicit per-publish approval to join a room and approval for every room_send, exposes joined rooms through status, and coalesces and caps incoming events so pages send summaries rather than streams.
• Data: Artifact MCP connector guidance — Expands Artifact MCP manifests from claude.ai and host connectors to include built-in connector guidance while retaining exact upstream-tool-name discovery.
• Data: Claude API reference — Python and Skill: Anthropic Python SDK 0.x to 1.x upgrade — Switch timeout and custom-client guidance to anthropic.Timeout/httpx2, expand output_format migration coverage, and preserve supported older-model sampling parameters through extra_body instead of deleting them blindly.
• Data: Background tasks changed event schema, SDK protocol capabilities field, Interrupt cancel queued parameter, Interrupt receipt still queued field, and SDK subagent stats schema — Add reinitialize snapshots for live background tasks, hosted-session exceptions to cancel-queued receipts, and held-back-result timing details for cost, duration, model usage, and subagent totals.
• System Prompt: Coordinator mode orchestration and Tool Description: ListAgents — Add capability-aware user-message routing, blocked worker outcomes, concise launch updates when no communications role exists, and teammates as addressable agents.
• System Reminder: Previously invoked skills — Broadens the post-compaction warning so request or argument text anywhere in restored skill bodies, including “User Request” sections, is historical context rather than a new live instruction.
• System Prompt: Interactive agent intro and System Prompt: Harness instructions — Add a collaborative-goals identity branch when no output style is active, alongside the existing software-engineering framing.
• Tool Description: Bash (Git commit and PR creation instructions) — Corrects generated pull-request bodies so summary and test-plan templates appear under their matching headings and optional attribution follows the test plan.
• System Prompt: Auto memory durable lesson instructions — Adds a runtime note about the configured memory directory alongside the persistent-memory introduction.

v2.1.238

2.1.238

(☝️ Click to see full diff.)

+3,292 tokens

• NEW: Data: SDK subagent stats schema — Documents cumulative per-session Agent-tool subagent counts by launch mode, type, nesting, outcome, and refusal, including reset, background-lifecycle, remote-launch, and result-stream timing caveats.
• NEW: Data: Self-hosted runner deferred shutdown timing notice — Explains deferred session release and fallback drain timing, supervisor stop-timeout sizing, forced-kill consequences, and second-signal behavior.
• NEW: Data: VCS state changed branch field — Defines the optional best-effort branch hint for commit and push events, including per-branch events for multi-branch pushes and uncertain-attribution cases.
• NEW: Tool Description: Artifact live room guidance — Adds transient, at-most-once rooms for live Artifact viewers, with untrusted event handling, bounded room_send broadcasts, peer-status results, and guidance to use republishes or the artifact database for durable state.
• NEW: Tool Description: Artifact runtime verification guidance — Adds post-publish verify diagnostics for console output, uncaught errors, failed resources, and capability calls, while treating viewer-reported diagnostics as untrusted data and no-viewer results as inconclusive.
• Data: Artifact MCP connector guidance — Allows supported sessions to expose locally configured MCP servers to Artifacts as host:<server>, while warning that they work only for viewers with the same local server connected.
• Data: Self-hosted runner command help — Adds command- and file-backed Proxy-Authorization header injection for egress proxies, including per-connection refresh, child-process proxy rewriting, secret-handling, and orchestrator limitations.
• Data: Self-hosted runner command help and System Prompt: Self-hosted runner doctor — Document --defer-shutdown-max-min, which stops new assignments while attached sessions continue until release or the ceiling, then parks or drains them under explicit signal, grace-period, and supervisor-timeout rules.
• Data: VCS state changed event schema — Clarifies that a push updating several branches emits one push event per branch.
• Skill: Artifact PR review and Skill: Artifact PR review (composed publish flow) — Harden decision-island extraction by anchoring searches to the full JSON script opening tag and reading from its end, avoiding collisions with prose that merely mentions the island ID.
• Skill: Artifact document — Stops requiring in-page status chips and author, date, or version metadata, treats those details as Artifact-owned chrome, and removes the related self-checks.
• Skill: Design — Expands canvas authoring with no-human-in-loop defaults, visible direction-selection artboards, Main.dc.html handoff rules, data-visualization routing, factual sample-value constraints, print dimensions, frame-fit and optional browser checks, image-basename and generic-name validation, template-binding and per-artboard-state caveats, static-artboard guidance, and required file_path, description, and favicon metadata on every publish.
• Skill: Prototype — Makes Artifact verify the sanctioned post-publish runtime check when available and warns that an empty no-viewer result does not prove the demo works.
• System Prompt: Artifact comment result guidance, Tool Description: Artifact comments guidance, and Tool Parameter: Artifact comment actions guidance — Require Claude activation for both replying to and resolving a comment thread, leaving non-activated threads for the commenter to resolve even after Claude addresses them elsewhere.
• System Prompt: Artifact comment thread framing — Adds untrusted multi-file page markers to identify which Artifact file a thread anchors to, and explicitly avoids assuming the main page when attribution is degraded.
• System Prompt: Coordinator mode orchestration — Adds capability-aware Skill-tool guidance alongside the existing workflow-tool notes.
• System Reminder: Artifact auto-replies resumed, Tool Parameter: Artifact watch actions guidance, and Tool Description: Artifact publishing and update guidance — Distinguish interruption pauses that keep the watch from killed or unwatched stops, answer comments sent during a pause after resuming, and recognize an “already registered” result as evidence of a remote watch.
• System Reminder: Output style active — Escapes the displayed active-style value as untrusted text and reads the turn reminder from the active style configuration.
• Tool Description: Artifact publishing and update guidance and Tool Description: Artifact runtime capabilities guidance — Document per-artifact browser storage for lightweight per-viewer conveniences, require failure-safe access, and prefer shared runtime capabilities for state that must be durable, cross-viewer, or readable by Claude.

v2.1.237

2.1.237

(☝️ Click to see full diff.)

+1,249 tokens

• NEW: System Prompt: Auto mode Slack message provenance — Treats bound-thread Slack relays that open with the server-verified human marker as direct user intent capable of clearing soft blocks, while bot-attributed or nested cross-session relays remain untrusted and cannot establish consent or launder permissions.
• NEW: System Prompt: Concise output style — Defines the built-in Concise style to lead with results, omit narration and repeated recaps, use short plain answers by default, and preserve requested detail and correctness, including full error reports, failing-test output, security warnings, and destructive-action confirmations.
• NEW: Tool Description: Poll — Adds an idle wait for queued harness events, returning pending events immediately and yielding to new user input, while defining authoritative envelope provenance, untrusted event content, nonce manifests, transcript-replay compatibility, and oldest-first chunked delivery.

v2.1.236

2.1.236

(☝️ Click to see full diff.)

+26,676 tokens

• NEW: Agent Prompt: Security monitor host-context line guidance and Data: Hook classifier context field — Distinguish live host-attached user statements, which may satisfy a soft-block consent bar, from restored or mixed host context, which remains unverified; define call binding, trust, size, timing, and rewrite-integrity rules, and integrate the guidance into the autonomous-action monitor.
• NEW: Data: SDK register device hooks request schema — Documents how cloud device clients register forwarded hooks and vetted worker-side templates, including the size limit, worker-epoch handling, and retryable versus terminal registration errors.
• NEW: Skill: Anthropic Python SDK 0.x to 1.x upgrade — Adds an executable migration workflow for the Python 3.10 floor, httpx2, removed deprecated APIs and parameters, raw-response and streaming changes, Bedrock region handling, verification, and user-owned migration decisions, with /claude-api upgrade routing and an authoritative migration-guide source.
• NEW: System Reminder: Artifact auto-replies resumed — Reports that a requested Artifact comment auto-reply resume is re-arming the live watch, explains which comments from the stopped period will be handled based on the stop cause, and warns that the stop persists until reconnection succeeds.
• NEW: System Reminder: Artifact type page untrusted content warning — Treats HTML supplied by an owned Artifact's type publisher as untrusted data that cannot grant instructions or permission escalation.
• NEW: System Reminder: Background task notification with concurrent user input — Separates an automated background-task event from genuine user input delivered in the same turn and prevents the notification from being treated as approval or consent.
• REMOVED: System Reminder: Ultrareview launch acknowledgement — Removes the standalone acknowledgement prompt for already-visible cloud review launches and remembered --fix intent.
• Agent Prompt: Managed Agents onboarding flow; Data: Managed Agents self-hosted sandboxes, memory stores reference, and related Managed Agents/AWS references — Expand self-hosted SDK-worker guidance with single-item handling, safe shutdown and file confinement, synchronized memory stores and recovery, resource and deployment constraints, and Claude Platform on AWS authentication, session-duration, and memory-store limitations.
• Data: Managed Agents tools and skills and related Managed Agents references — Add per-tool web_search and web_fetch domain filters, search location and fetch-size settings, validation and runtime behavior, typed-SDK configuration changes, multiagent restriction layering, and guidance that these tools run server-side rather than under environment networking rules.
• Data: Managed Agents events and steering and Data: Managed Agents overview — Document the Console session viewer's searchable transcript, timeline, raw events, tool/resource/thread inspection, cost views, JSON export, and event deep links.
• Data: Plugin eval and skill-doctor reference — Requires every eval case to provide an execution prompt, treating it as the resumed session's next user turn when a history file is also supplied.
• Skill: Artifact document — Clarifies that publishing assigns comment-anchor block IDs, the editor assigns IDs to user-added blocks, and existing IDs must still be preserved rather than copied or hand-authored unnecessarily.
• System Prompt: Artifact comment reply composer — Makes change-request reply guidance conditional instead of always prescribing a generic work-in-progress acknowledgement, while retaining brief plain-text answers and prohibitions on claiming completed edits.
• Tool Description: Artifact database guidance — Adds file-backed database reads via out_dir and writes via a local JSON file_path so large or numerous documents need not be returned or retyped inline.
• Tool Description: Artifact publishing and update guidance and Tool Parameter: Artifact watch actions guidance — Add terminal and web paths for reopening artifacts, distinguish watch arming from an established connection, limit comment wakes to armed auto-replies, restore at most one watch after resume or continue, and tighten status-based claims about active watches.
• Tool Description: Edit, Tool Description: Edit single replacement, and Tool Description: Write — Make read-before-edit and read-before-overwrite guidance path-sensitive, explicitly retaining the prior-Read requirement for files outside the working directory.
• Tool Description: SendMessage cross-session guidance — Adds one-shot notify_when_idle subscriptions for local sessions, including pure subscriptions, approval-held notices, expiry behavior, and guidance to use them instead of polling or status-chasing messages.

v2.1.235

2.1.235

(☝️ Click to see full diff.)

+6,990 tokens

• NEW: Data: SendMessage ambiguous recipient display — Adds user-facing “not sent” explanations for inexact or duplicate agent names, incomplete session searches, and recipients that need exact-name or pinned-identity confirmation.
• NEW: System Prompt: Artifact comment fast acknowledgement selection — Selects one canned pre-reply acknowledgement from editability, trigger history, and whether the newest comment asks for a specific edit, broader inspection, or a thread-only answer, with a safe ambiguity and off-topic fallback.
• NEW: System Prompt: Non-fork subagent delegation examples — Consolidates synchronous and background delegation examples into capability-aware guidance that suppresses default-agent examples when general-purpose agents are unavailable.
• NEW: System Reminder: Ultrareview launch acknowledgement — Briefly acknowledges an already-visible cloud review without repeating its details, preserves --fix intent, and relates later findings to review notes the cloud pass cannot see.
• REMOVED: System Prompt: Background subagent delegation examples and System Prompt: Foreground subagent delegation examples — Remove the separate execution-mode variants superseded by the capability-aware non-fork delegation prompt.
• Data: Plugin eval and skill-doctor quick reference, Data: Plugin eval and skill-doctor reference, and Skill: Plugin eval authoring interview — Promote configurable --eval-dir suites from upcoming to current behavior, document containment-checked plugin discovery and result placement, teach the interview to honor custom suite paths and treat plugin paths as data, and distinguish tool-free negative checks from file-content graders that require an existing file.
• Skill: Artifact document — Adds block-ID rules that keep viewer comments anchored, preserve editor-generated IDs across edits, strip IDs from duplicated blocks, and reserve hand-written IDs for short in-page link targets.
• System Prompt: Forked agent guidance and Tool Description: Agent (when to launch subagents) — Make omitted subagent_type behavior conditional on general-purpose-agent availability, add plan-specific subagent restrictions, and provide fallback instructions when the general-purpose type is unavailable.
• Tool Description: Artifact database guidance — Updates current-viewer identity lookup from claude.user.id() to id() on the page’s user capability.

v2.1.234

2.1.234

(☝️ Click to see full diff.)

+11,405 tokens

• NEW: Data: SDK cloud session init snapshot field, Data: SDK control cancel request schema, Data: SDK request user dialog kind field, and Data: SDK result message schema — Document cloud-session attachment and directory-sync state, cancellation of in-flight control requests, capability-negotiated user-dialog kinds, and the turn-complete result contract.
• NEW: Data: syncClaudeAiSkills setting — Documents how false disables account-synced skills across user, managed, workspace, and invocation scopes, including hiding, trash cleanup, re-enablement, and unsupported project-setting behavior.
• NEW: Data: VCS state changed event schema — Restores the best-effort repository cache-invalidation event with working-directory hints, stricter push-branch attribution, and caveats for silenced or backgrounded mutations.
• NEW: Skill: Claude guide unavailable reference fallback — Adds an embedded live-source index and WebFetch fallback when the Claude guide's on-disk reference files cannot be written.
• NEW: Tool Description: Artifact assets guidance — Adds upload, list, read, reference, and permanent-delete guidance for files in artifacts that declare the assets capability.
• NEW: Tool Parameter: Artifact watch actions guidance — Adds action-level guidance for watch, unwatch, status, durable remote wakes, comment wakes, and explicitly authorized resume_replies behavior.
• REMOVED: Skill: Build with Claude API (reference guide) — Removes the redundant standalone quick-task navigation prompt; the primary Claude API skill retains its integrated reading guide.
• REMOVED: System Reminder: File modification detected (budget exceeded) and System Reminder: File modified by user or linter — Remove the standalone reminders for externally modified files and omitted modification snippets.
• REMOVED: System Reminder: MCP resource no displayable content — Consolidates the no-displayable-content case into the generalized MCP resource status reminder.
• Agent Prompt: Coding session title generator and Agent Prompt: Session title and branch generation — Shift titles to specific, two-to-five-word noun phrases that omit generic task verbs and abstract action labels, retain recognizable identifiers, and follow the user's language while keeping branch names in English.
• Agent Prompt: Status line setup — Adds GitLab merge-request metadata and labeling alongside GitHub pull-request status-line support.
• Data: Artifact decision component script, Data: Workshop artifact HTML template, and Skill: Artifact components — Harden verifier-pinned decision scripts against ambiguous script tags, quoted-attribute boundaries, escaped script states, and non-ASCII whitespace while refreshing the blessed digest.
• Data: Plugin eval and skill-doctor quick reference and Data: Plugin eval and skill-doctor reference — Promote image judging to current behavior, document binary-grading remedies, strengthen sandbox isolation from host repositories and project settings, and add plugin-load advisories, authentication-partial results, richer quiet-JSON diagnostics, and SIGTERM handling.
• Skill: Plugin eval authoring interview — Makes tools and execution budgets follow each grader's actual side effects, rejects pilots whose plugin failed to load or whose graders cannot pass with granted tools, and adds image and binary-artifact grading guidance.
• Skill: Building LLM-powered applications with Claude — Clarifies that cited language and shared reference paths are external skill files that must be read on demand before relying on them.
• Skill: Artifact design, Skill: Design, Skill: Prototype, and Tool Description: Artifact publishing and update guidance — Allow Google Fonts stylesheets and font files as the sole external-host exception, require fallback stacks, and note export fallbacks where applicable.
• Skill: Artifact document, Skill: Artifact slides, and Skill: Artifact spreadsheet — Switch collaborative editing to explicit whole-artifact saves under the artifact capability, distinguish write-access and view-only behavior, preserve reader changes through conflict-safe rereads, remove obsolete server-owned data-id restrictions, support structural slide edits, and make spreadsheet scratch cells persist when saved.
• Skill: Artifact PR review and Skill: Artifact PR review (composed publish flow) — Make decision controls the default unless a display-only page was requested, and clarify that connector-backed live PR data—not artifact saving alone—is what restricts external sharing.
• Skill: Design — Pins every canvas publish to runtime contract 0.1.31, asks whether app concepts should be static or clickable, leaves publish approval to the tool and handles refusals without unsafe retries, preserves existing capability declarations on updates, clarifies that export—not saving—limits sharing to the organization, and favors a concise handoff followed by a content-focused second pass.
• System Prompt: Artifact comment list framing and System Prompt: Artifact comment thread framing — Add trusted attribution for comments posted through an artifact's own interface, treating sent comments as the account holder's request while asking when they conflict with directly typed feedback.
• System Prompt: Coordinator mode orchestration — Adapts peer-tool and post-launch instructions to available capabilities and identifies worker results inside harness-generated system reminders without reproducing their wrapper or XML.
• System Reminder: Compact file reference, System Reminder: File opened in IDE, System Reminder: File truncated, System Reminder: Large PDF read guidance, and System Reminder: Lines selected in IDE — Escape untrusted filenames before interpolation and soften the instruction about mentioning file truncation.
• System Reminder: MCP resource no content — Escapes server and URI attributes and generalizes the rendered resource status to cover both no-content and no-displayable-content cases.
• Tool Description: Artifact comments guidance and Tool Description: Artifact publishing and update guidance — Detect when remote artifact watching is unavailable, avoid claiming an active watch, and direct users to claude --watch-artifact <url> on their own machine.

v2.1.233

2.1.233

(☝️ Click to see full diff.)

+27,728 tokens

• NEW: Data: Plugin eval and skill-doctor quick reference and Data: Plugin eval and skill-doctor reference — Add condensed and comprehensive offline guidance for early-access claude plugin eval, eval init, and /skill-doctor, covering enablement, suite authoring, graders, run options, result and report formats, sandboxing, CI, and troubleshooting.
• NEW: System Prompt: Plugin eval enabled-session status — Announces when plugin eval is enabled and gives the exact CLAUDE_CODE_WALNUT_SPIRE=1 fallback for clients and CI that cannot receive the organization rollout, with supported shell, user-settings, and managed-settings locations and a warning not to rely on project settings.
• Agent Prompt: Claude Code guide, Agent Prompt: Claude guide agent, Data: Claude Code live documentation sources, Data: Claude Code recent changes reference, and Skill: Claude Code configuration guide — Route plugin-evaluation and skill-diagnostics questions through current-build checks and the new offline references, distinguish /skill-doctor from linting, and warn against stale-memory answers, guessed documentation URLs, or invented enablement variables.
• Data: Artifact decision component script, Data: Workshop artifact HTML template, and Skill: Artifact components — Update decision controls to acquire artifact publishing through the asynchronous viewer 0.2 claude.use('artifact') API, retain viewer 0.1 artifact/self compatibility, arm only after capability availability, and refresh the verifier-pinned digest.
• Data: Claude Code gateway protocol — Relays Anthropic 400/413 error messages needed for client recovery such as auto-compaction while continuing to sanitize other upstream messages and preserve error types.
• Skill: Artifact document — No longer presents document artifacts as supporting selection-based comments or requires hidden comment-store machinery, while preserving live editing and reframing review language around feedback.
• Skill: /doctor slash command and Skill: /doctor slash command description — Add diagnosis of malformed skill YAML frontmatter, explaining that parse failures drop every field and trigger fallback naming and descriptions while silently disabling tool, model, and invocation settings.
• Skill: Plugin eval authoring interview — Keeps every calibration pilot and re-pilot private with --no-publish and tells users how to keep final full-suite reports local.
• System Reminder: Team Coordination and Tool Description: SendMessage — Make task-list resources and coordination instructions conditional on available task tooling and allow legacy status updates in plain prose when those tools are absent.
• Tool Description: WebFetch, Tool Description: WebFetch (concise), and Tool Description: WebFetch private URL warning — Derive cache-expiry text at render time instead of hard-coding 15 minutes.

v2.1.232

2.1.232

(☝️ Click to see full diff.)

+48,736 tokens

• NEW: Agent Prompt: Web fetch agent usage guidance and Agent Prompt: Web reading specialist — Add a dedicated WebFetch delegation flow that returns focused, source-grounded reports from untrusted pages, supports follow-up questions about already-read content, and confines binary-file handling to harness-reported tool-results paths.
• NEW: Skill: Artifact components and Data: Artifact decision component assets — Add reusable, verifier-pinned decision blocks for non-workshop HTML artifacts, including canonical design tokens, styles, markup, and scripts for persisted selections and readback, plus composition and injection-safety constraints.
• NEW: System Prompt: Artifact comment fast acknowledgement — Adds a no-tools, single-sentence acknowledgement under 160 characters before the full comment response, distinguishing change requests from questions while preserving plain-text and internal-handling restrictions.
• NEW: System Reminder: Bound conversation activity authority warning — Treats bound-conversation edits and reactions as awareness-only, never as fresh instructions, approval, consent, or a way around a denial, while still allowing relevant activity to inform work in progress.
• NEW: Tool Description: Background monitor push notification guidance — Directs background monitors to push only events that materially change what the user should do next, such as a new error or a status transition they were awaiting.
• REMOVED: Agent Prompt: /code-review workflow routing and System Prompt: Code review artifact publishing instructions — Remove the standalone prompts for routing /code-review through a background workflow and publishing its findings as a shareable Artifact.
• REMOVED: Agent Prompt: WebFetch summarizer — Removes the inline page-content summarizer superseded by the dedicated web-reading agent flow.
• REMOVED: Data: VCS state changed event schema — Removes the standalone schema for best-effort repository-state cache-invalidation events emitted after detected foreground VCS mutations.
• Agent Prompt: Security monitor for autonomous agent actions (second part) — Extends real-browser protections to Chrome tools reached through the remote-device bridge and hard-blocks attacks on recognizable third-party systems outside the task's trust boundary unless an exercise or authorized engagement designates the target.
• Agent Prompt: Worker fork — Updates the fork agent's availability description from the “fork experiment” to the “fork gate.”
• Data: Managed Agents multiagent sessions — Removes the temporary exclusion of Fable advisors so valid pairings mirror the Messages advisor-tool pairing table.
• Data: Workshop artifact HTML template; Skill: Artifact PR review, Skill: Artifact PR review (composed publish flow), Skill: Design, and Skill: Whiteboard — Migrate self-update guidance and clients to the artifact capability spelling while retaining legacy self compatibility, and clarify that design canvases open ready to edit but cannot retain changes when artifact publishing is unavailable.
• Skill: Artifact design and Tool Description: Artifact — Require design calibration before writing both HTML and Markdown artifacts, treating format as a deliberate choice and forbidding Markdown as a speed shortcut while preserving the workshop-specific exceptions.
• Skill: Prototype — Adds logic-first prototypes with full-state walkthroughs for behavior questions, requires every prototype to state one design question, verifies source-derived shell bytes against trusted registry digests before reuse, and keeps structurally distinct exploratory variants in one artifact until a direction is chosen.
• System Prompt: Artifact comment edit composer — Aligns edit replies with the shared plain-text formatting and internal-handling nondisclosure restrictions used by comment replies and fast acknowledgements.
• System Reminder: Artifact comment reply activation failure and Tool Parameter: Artifact comment actions guidance — Clarify that comment-thread activation survives artifact republishes and renames, while deactivation or thread deletion can clear it.
• Tool Description: ListAgents and Tool Description: SendMessage cross-session guidance — Clarify that exact live names deliver across local, remote, and cloud sessions, that references are only for ambiguity or lookup failures, and that cloud sessions can receive messages but cannot yet reply to another session.
• Tool Description: SendFeedback drafting guidance — Tightens feedback privacy by replacing personal identifiers with roles, excluding customer channel IDs and excerpts, constraining file-path evidence, and describing suspected vulnerabilities without working exploits or extraction steps.

v2.1.231

2.1.231

(☝️ Click to see full diff.)

_{No changes to the system prompts in v2.1.231.}