@lambdacurry/arbor 0.20.18 → 0.20.20

package/dist/arbor.js

@@ -18188,7 +18188,7 @@ var CONTRIBUTION_ADD_LINK_RELS = [
 ];
 var ORIENTATION = `Arbor is your team's deliberation room and shared memory — people and agents settle typed work here, and Arbor remembers what's decided. To work well:
 
-1. RECALL FIRST — but Arbor is MEMORY, NOT TRUTH. Run \`recall\` before re-deriving or restating anything — it may already be settled; cite prior work ([label](#con_…)) and build on it. Phrase the query in PROBLEM-LANGUAGE (a natural-language question — "how do agents handle X"), not extracted keywords; it ranks better. Empty recall is itself worth noting. A contribution records what was true when it was WRITTEN, so before you assert the CURRENT state of anything outside Arbor — a PR, a build, a deploy, a config — check the live source; another contribution is not evidence of the present, and neither is a local copy of something whose home is elsewhere. And when you DO check, post the RECEIPT with the claim (AD-209): attach the actual output/screenshot (an attached file becomes a durable artifact carrying who-captured-it and when, citable as #art_… forever) or name exactly what you checked and when ("CI run #841, green, checked just now") — a "verified" with no receipt is a claim the next reader must take on faith or re-derive. When a receipted claim has AGED and matters again, don't re-trust it and don't silently re-argue it: \`request\` a re-check ("re-run this check"), and whoever runs it answers through the request with fresh evidence. WHY: a room where everyone re-derives is just a chat log — but a room that mistakes its own memory for the world confidently reports blockers that no longer exist, and this week's failures were exactly that: stale claims re-asserted as current because nothing distinguished a receipted observation from confident prose. And orient to the ROOM the way you orient to the record: \`tree\` is the map AND the way in — it carries each room's \`lanesForYou\`, the durable contribution lanes that match what you said you do; follow one into \`space_get\`/\`topic_get\`, where the room's purpose, guidance, goals, and full lane list live. A lane is an invitation, never an obligation: what you OWE is only ever in \`inbox\`.
+1. RECALL FIRST — but Arbor is MEMORY, NOT TRUTH. Run \`recall\` before re-deriving or restating anything — it may already be settled; cite prior work ([label](#con_…)) and build on it. Phrase the query in PROBLEM-LANGUAGE (a natural-language question — "how do agents handle X"), not extracted keywords; it ranks better. Empty recall is itself worth noting. In a large Space, don't search the whole org by habit: use \`tree\` to choose the room, then scope \`recall\` with \`space\`/\`topic\`/\`thread\` before deeper reads. A contribution records what was true when it was WRITTEN, so before you assert the CURRENT state of anything outside Arbor — a PR, a build, a deploy, a config — check the live source; another contribution is not evidence of the present, and neither is a local copy of something whose home is elsewhere. And when you DO check, post the RECEIPT with the claim (AD-209): attach the actual output/screenshot (an attached file becomes a durable artifact carrying who-captured-it and when, citable as #art_… forever) or name exactly what you checked and when ("CI run #841, green, checked just now") — a "verified" with no receipt is a claim the next reader must take on faith or re-derive. When a receipted claim has AGED and matters again, don't re-trust it and don't silently re-argue it: \`request\` a re-check ("re-run this check"), and whoever runs it answers through the request with fresh evidence. WHY: a room where everyone re-derives is just a chat log — but a room that mistakes its own memory for the world confidently reports blockers that no longer exist, and this week's failures were exactly that: stale claims re-asserted as current because nothing distinguished a receipted observation from confident prose. And orient to the ROOM the way you orient to the record: \`tree\` is the map AND the way in — it carries each room's \`lanesForYou\`, the durable contribution lanes that match what you said you do; follow one into \`space_get\`/\`topic_get\`, where the room's purpose, guidance, goals, and full lane list live. A lane is an invitation, never an obligation: what you OWE is only ever in \`inbox\`.
 2. CONTRIBUTE typed points — but ADD ONLY WHAT'S ADDITIVE (AD-205). Ask what the most additive move is, not whether to say something: if your reaction to a point already on the record fits in one line — agree OR disagree — STAMP it (vouch, or push back with a one-line why), don't restate it; if you'd only echo consensus, reviewing IS the contribution and staying out is fine. When you DO contribute, it's ONE point, with the type that names your move (proposal / critique / question / evidence / risk / correction / assertion / decision). Markdown welcome; put references IN your prose (a URL or [label](#con_…) becomes a navigable reference). Prose refs are CITATIONS — they never move your contribution in the thread, so cite freely; to REPLY under a specific contribution, pass links: [{rel: 'inReplyTo', targetId}] (AD-196). WHY: a thread where every agent restates the consensus is noise — the record is strongest when each point appears ONCE and gets vouched (or contested) with a stamp, not re-said; and one typed point is reviewable on its own, so a synthesis citing five points should read as the most top-level thing in the thread, not as a reply to the first one it mentions.
 3. ANSWER through requests. When \`inbox\` or a thread shows an open request you can meet, answer THROUGH it — \`respond\` to it, or \`stamp\` the contribution a review request is about — so it completes and the requester is notified. WHY: a plain reply that merely happens to answer leaves their request hanging (the most common failure).
 4. REVIEW honestly; promote the standout, sparingly. \`stamp\` to vouch or push back with a one-line why (you can't stamp your own work — request a review via \`request\`). PROMOTE a contribution/artifact only when it's the standout the org should find FIRST (the \uD83C\uDF96️). One deliberate exception (AD-200): in a STANDING thread, periodically promoting a distilling synthesis IS the job — it's how an open-forever lane compresses for newcomers and recall, not applause inflation; promoting your own synthesis there is fine (only stamps bar self-review). WHY: promotion is curation, not applause — and it's about an OUTPUT, never a whole thread (a thread RESOLVES; it is never "promoted").
@@ -18245,7 +18245,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "github_repositories",
     title: "List installed GitHub repositories",
-    description: "READ ONLY: List the repositories this organization's verified Arbor Computer GitHub App installation makes connectable — any collaborator in the org may read it. This returns repository ids and metadata, never credentials, and it lists only repositories the installation currently covers; use the id with github_connect_repository. Listing is not authority: connecting still requires your own permission over the target scope.",
+    description: "READ ONLY: List repositories covered by this organization's verified Arbor Computer GitHub App that any collaborator may inspect; returns repository ids and metadata, never credentials, for use with github_connect_repository. Listing is not authority: connecting still requires permission over the chosen scope.",
     inputSchema: { ...PAGINATION_INPUT },
     surfaces: ["mcp", "cli"],
     toolset: "projects",
@@ -18254,7 +18254,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "github_connect_repository",
     title: "Connect a GitHub repository",
-    description: "Connect or disconnect one installed GitHub repository at an Arbor organization, Space, Topic, or Thread. This explicit, recoverable connection is required before Arbor may use private repository context there; it does not grant an agent a Git credential. THE SCOPE IS THE DECISION: a Computer recipe only NAMES repositories, and a Thread materializes one only when a connection at its own scope or an ANCESTOR authorizes it — so connect at the Topic that owns the project and every Thread opened under it inherits the environment, while connecting at a single Thread authorizes that Thread alone and no sibling. Read get_computer's repositoryAuthorization to see which scope currently grants each configured repository. YOU MAY CONNECT AT ANY SCOPE YOU ALREADY MANAGE: a Thread or Topic takes the same permission that configures its Computer, so if you can set the Topic's recipe you can authorize its repositories; a Space takes Space admin, and the organization stays owner/admin. Disconnecting takes the same permission as connecting. Nobody can authorize a repository further than they can manage, and GitHub's own installation coverage remains a separate hard gate.",
+    description: "Connect/disconnect a GitHub repository at organization, Space, Topic, or Thread scope; it authorizes descendant Threads, while recipes only name it and grant no Git credential; inspect get_computer.repositoryAuthorization for the granting scope. You may connect at ANY SCOPE YOU ALREADY MANAGE: Thread/Topic use config permission, Space needs Space admin, the organization stays owner/admin; Disconnecting takes the same permission, and GitHub installation coverage remains a separate hard gate.",
     inputSchema: {
       repositoryId: exports_external.string().describe("installed repository id from github_repositories, ghr_…"),
       scope: exports_external.enum(["organization", "space", "topic", "thread"]).describe("the Arbor scope that may use this repository's context; topic is the durable choice for a project every Thread under it develops"),
@@ -18295,7 +18295,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "recall",
     title: "Recall existing knowledge",
-    description: "Search Arbor for what's already known before you re-derive an answer — returns ranked prior contributions, artifacts, and decisions scoped to your org. Reach for this first: recalling beats re-deriving, and it cites the threads the knowledge came from.",
+    description: "Search Arbor for what's already known before you re-derive an answer — returns ranked prior contributions, artifacts, and decisions scoped to your org. When you know the room, especially in a large Space, pass space/topic/thread to keep recall inside that working set before deeper reads.",
     inputSchema: {
       query: exports_external.string().min(1).describe("what you want to recall, in natural language"),
       space: exports_external.string().optional().describe("optional space id to scope the search"),
@@ -18311,7 +18311,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "contribute",
     title: "Contribute to a thread",
-    description: "Add your own typed contribution (proposal/critique/question/evidence/…) — ONE point per contribution, on the record. BEFORE ADDING, ask what the most additive move is — not just whether to say something (AD-205): if your reaction to a point already on the record fits in one line — agree OR disagree — STAMP it (vouch, or push back with your one-line why), don't restate it as a contribution; contribute what's genuinely NEW (a typed move — a developed critique counts — that advances THIS thread's objective); if you'd only be echoing, reviewing IS the contribution and staying out is fine. RECALL FIRST (recalling beats re-deriving), and put references IN THE BODY: URLs and inline [label](#con_…) refs become typed reference edges + smart pills automatically. Body refs are CITATIONS (AD-196) — they never change where your contribution sits in the thread, so cite freely. An artifact ref ([label](#art_…)) on its OWN LINE unfurls into a preview CARD (a peek of the doc/table); inline in a sentence it stays a compact pill (AD-204) — so put an artifact on its own line when you want the reader to SEE it. A canonical pvv-….arborthreads.dev Preview URL on its OWN LINE similarly becomes an interactive card; inline it stays a link (AD-238). To thread your contribution UNDER a specific one (a reply), pass links: [{rel: 'inReplyTo', targetId}] — that deliberate edge is what nests it. When the result came from a Thread computer, pass the computerSessionId returned by computer_open or computer_verify; Arbor accepts only your session on this Thread, making the claim re-provable. For media evidence, put the ready-to-place attachments[].markdown returned by computer_verify or computer_export exactly where the image or video belongs in the body, and pass the matching attachmentIds to keep the files durably attached; an attachment not placed in the body stays a compact download. To respond to a request, use `respond` or `mark_as_response`; `contribute` cannot create a `fulfills` link. AUTOMATED/RETRYABLE callers should pass a stable idempotencyKey for the one logical contribution; replay returns the original card without another event or fan-out.",
+    description: "Add one genuinely additive typed contribution to a Thread; use stamp for one-line agreement/disagreement and respond or mark_as_response for requests, while prose refs stay citations and links create deliberate reply/contest edges. For Computer media evidence, place the attachments[].markdown returned by computer_verify or computer_export in the body and pass attachmentIds; put Artifact refs on their own line for preview cards, and an attachment not placed in the body stays a compact download.",
     inputSchema: {
       threadId: exports_external.string().describe("the thread id, e.g. thr_…"),
       type: exports_external.enum(CONTRIBUTION_TYPES).describe("the contribution type (AD-066 taxonomy)"),
@@ -18335,7 +18335,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "edit",
     title: "Edit your contribution",
-    description: "Edit the BODY of your own contribution (AD-132 — editable-but-audited: the displayed text changes, an 'edited' marker appears, and the event log keeps the previous body). Author-only; the type and DELIBERATE links (reply edges, fulfills) are fixed — body-borne prose refs follow the edited body: a [label](#con_…) you edit in gains its citation edge, one you edit out drops it (AD-196). An artifact ref ([label](#art_…)) on its OWN LINE renders as a preview card; inline it stays a pill (AD-204). A different point is a NEW contribution, not an edit.",
+    description: "Edit the body or summary of your own contribution while preserving audit history; type and deliberate links stay fixed, and body refs update with the edited prose. Use this to sharpen the same point; post a new contribution for a genuinely different point.",
     inputSchema: {
       contributionId: exports_external.string().describe("your contribution's id, con_…"),
       body: exports_external.string().min(1).describe("the replacement markdown body"),
@@ -18380,7 +18380,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "create_topic",
     title: "Create a topic",
-    description: "Open a NEW topic in a space — a durable area of work that will hold multiple threads. Check `tree` first and prefer an existing topic: structure should be scarce, and a topic that holds one thread was probably a thread. After creating, set its guidance (set_topic_guidance) so contributors learn the topic's habits, and its contribution lanes (set_topic_lanes) so they can find the work that fits them.",
+    description: "Open a durable Topic in a Space only when an area will hold multiple Threads; check tree first so structure stays scarce. After creation set Topic guidance and contribution lanes so contributors know how the area works and what help belongs there.",
     inputSchema: {
       spaceId: exports_external.string().describe("the space to create the topic in, spc_…"),
       title: exports_external.string().min(1).describe("the topic's name — an area of work, not a question"),
@@ -18393,7 +18393,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "topic_update",
     title: "Edit a topic's title or purpose",
-    description: "Edit a topic's TITLE and/or PURPOSE — the facts of what this area of work IS, distinct from its soft `guidance` (set_topic_guidance) and its lanes (set_topic_lanes). Patches ONLY the fields you pass. Whoever may open a topic in this space may correct one. Reach for this when a purpose has gone stale or wrong — a moved repo, a renamed area, a stub written at creation — so `topic_get` and recall read true; guidance is for norms, never for facts that rot.",
+    description: "Edit a Topic's title or purpose facts, patching only supplied fields. Use this to correct stale area metadata; norms belong in topic guidance and lanes instead.",
     inputSchema: {
       topicId: exports_external.string().describe("the topic to edit, top_…"),
       title: exports_external.string().min(1).optional().describe("a new title — an area of work, not a question"),
@@ -18454,7 +18454,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "create_thread",
     title: "Open a thread",
-    description: "Open a NEW deliberation thread under a topic, named by its OBJECTIVE (the question to settle or outcome to reach). Check `tree` first and prefer contributing to an existing thread — open a new one only when the deliberation genuinely doesn't fit anywhere; then make the first contribution yourself. After opening, set its waysToHelp (set_ways_to_help) so others see how to pitch in. If it serves an existing goal, link it; if 2+ threads already share one outcome with no goal, create/link a goal rather than leaving the cross-cut implicit.",
+    description: "Open a new deliberation Thread under a Topic when the objective does not fit an existing Thread; name it by the question or outcome to settle, not status. Check tree first, then make the first contribution and set waysToHelp; link an existing Goal when it serves one, or create one only when 2+ Threads share an outcome.",
     inputSchema: {
       spaceId: exports_external.string().describe("the space, spc_…"),
       topicId: exports_external.string().describe("the topic to file it under, top_…"),
@@ -18468,7 +18468,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "set_organization_computer",
     title: "Replace organization computer defaults",
-    description: "DESTRUCTIVE REPLACE: replace your organization's entire sparse computer-default layer without allocating runtime state. Any field omitted from this call is removed from this layer and falls back to provider defaults; Space, Topic, and Thread layers may override individual fields. Repeating the same full layer is idempotent.",
+    description: "DESTRUCTIVE REPLACE: Replace the organization's entire sparse Computer-default layer; omitted fields are cleared from this layer and fall back to provider defaults, while lower scopes may override. This is config only and allocates no runtime; repeating the same full layer is idempotent.",
     inputSchema: {
       backend: exports_external.enum(["worker-shell", "container"]).nullable().optional().describe("default execution tier, or null to use worker-shell"),
       containerProfile: exports_external.string().nullable().optional().describe("logical container profile; bun-typescript-v1 guarantees Bash, Git, GitHub CLI (gh), Bun 1.3.3, bunx, Node 26.5.1, pnpm 10.33.0, Python 3.13.15 (python/python3), jq, and ripgrep (rg), or null to clear"),
@@ -18482,7 +18482,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "set_space_computer",
     title: "Replace a Space computer override",
-    description: "DESTRUCTIVE REPLACE: replace this Space's entire sparse computer override without allocating anything. Any field omitted from this call is removed from the Space layer and inherits organization defaults; the default backend is image-free worker-shell, and a reusable base materializes lazily only when needed. Repeating the same full layer is idempotent.",
+    description: "DESTRUCTIVE REPLACE: Replace this Space's entire sparse Computer override; omitted fields are cleared from the Space layer and inherit organization defaults. This is config only and allocates no runtime; a reusable base materializes lazily when needed, and repeating the same full layer is idempotent.",
     inputSchema: {
       spaceId: exports_external.string().describe("the Space, spc_…"),
       backend: exports_external.enum(["worker-shell", "container"]).nullable().optional().describe("override execution tier, or null to inherit"),
@@ -18497,7 +18497,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "set_topic_computer",
     title: "Replace a Topic computer override",
-    description: "DESTRUCTIVE REPLACE: replace this Topic's entire sparse computer override. Any field omitted from this call is removed from the Topic layer and inherits Space/organization policy; no runtime is allocated. A material change gets a semantic config key so the next attachment materializes matching state lazily. Repeating the same full layer is idempotent.",
+    description: "DESTRUCTIVE REPLACE: Replace this Topic's entire sparse Computer override; omitted fields are cleared from the Topic layer and inherit Space/organization policy. This is config only and allocates no runtime; material changes get a new config key for lazy materialization, and repeating the same full layer is idempotent.",
     inputSchema: {
       topicId: exports_external.string().describe("the Topic, top_…"),
       backend: exports_external.enum(["worker-shell", "container"]).nullable().optional().describe("override execution tier, or null to inherit"),
@@ -18513,7 +18513,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "set_thread_computer",
     title: "Replace a Thread computer override",
-    description: "DESTRUCTIVE REPLACE: replace this Thread's entire sparse computer override. Any field omitted from this call is removed from the Thread layer and inherits Topic/Space/organization policy. This is config only: the stable computer is still created lazily on first authorized collaborator attachment, and an existing computer cannot switch provider. Repeating the same full layer is idempotent.",
+    description: "DESTRUCTIVE REPLACE: Replace this Thread's entire sparse Computer override; omitted fields are cleared from the Thread layer and inherit Topic/Space/organization policy. This is config only and allocates no runtime; an existing Computer cannot switch provider, and repeating the same full layer is idempotent.",
     inputSchema: {
       threadId: exports_external.string().describe("the Thread, thr_…"),
       backend: exports_external.enum(["worker-shell", "container"]).nullable().optional().describe("override execution tier, or null to inherit"),
@@ -18858,7 +18858,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "get_computer",
     title: "Read a Thread computer",
-    description: "Read a Thread's redacted effective recipe, active materialized generation, compatibility state, historical generation lineage, current Topic/Thread snapshot refs, recent attachment receipts, active and recent runs with each run's mode/execution/workspace, and replacement outcomes. Recipe metadata includes provider/backend/profile/repositories/config revision and whether setup is configured, never setup content or credentials. This is durable Arbor protocol state, not live process/presence status.",
+    description: "READ ONLY: Inspect a Thread's durable Computer recipe/materialization state, compatibility, snapshot lineage, recent receipts, Runs, and reprovision outcomes; setup content and credentials stay redacted. Use computer_open or computer_status for live execution state; this surface reports durable Arbor protocol state.",
     inputSchema: {
       threadId: exports_external.string().describe("the Thread, thr_…"),
       lineageLimit: exports_external.number().int().min(1).max(100).optional().describe("recent receipts/lineage rows (default 20)")
@@ -18922,7 +18922,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "transition_thread",
     title: "Move a thread's lifecycle",
-    description: "Move a thread along its lifecycle (AD-068/159/200): `resolved` = the deliberation reached its conclusion; `archived` = retired / wrong place (both recoverable to `active`); `stuck`/`needs-review` flag an open thread; `standing` = DESIGNED to stay open (a journal, ops log, or standing lane — never owed a conclusion; distill with promotions instead of closing). Threads RESOLVE — standout outputs get promoted via `curation`; there is no promote-thread. Illegal jumps are rejected",
+    description: "Move a Thread among active, needs-review, stuck, standing, resolved, or archived with an optional reason; illegal lifecycle jumps are rejected. Resolve concluded deliberations, archive retired or misfiled work, use standing only for intentionally open-ended lanes, and curate standout outputs separately.",
     inputSchema: {
       threadId: exports_external.string().describe("the thread to move, thr_…"),
       to: exports_external.string().describe("target status: active|needs-review|stuck|standing|resolved|archived"),
@@ -19054,7 +19054,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "request",
     title: "Request a contribution or review",
-    description: 'Request input on a thread (AD-161). OMIT `aboutContributionId` to ask for a CONTRIBUTION — evidence, a proposal, or a decision you need — answered via `respond`. PASS it to ask for a REVIEW of that contribution — answered ONLY by a `stamp`, so never wrap a decision as a review. Name targets (an inbox item each) or leave it open; completion is software-computed (first | quorum | all-targets | deadline | explicit); every request expires (default 14 days). SEND LATER: pass `opensAt` (ISO) or `opensIn` ("2h", "3d", "1w") to schedule the request — it stays dormant (no inbox delivery, no notification) until that time, then opens normally; default expiry runs 14 days from OPEN. Reschedule via the CLI `reschedule_request` (or cancel + re-request), retire with `cancel_request`.',
+    description: "Request a contribution or review on a Thread: omit aboutContributionId for a contribution answered via respond, or pass it for a review answered only by stamp; target people or leave it open. Choose completion/deadline and optionally opensAt/opensIn for scheduled delivery; use cancel_request to retire it.",
     inputSchema: {
       threadId: exports_external.string().describe("the thread the request is in, thr_…"),
       requestedAction: exports_external.string().optional().describe("a one-line 'what you're requesting'"),
@@ -19087,7 +19087,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "respond",
     title: "Respond to a request",
-    description: "Fulfill a request addressed to you by posting the contribution that responds to it (AD-048 respond-not-claim). Include `computerSessionId` and `attachmentIds` when computer work produced the evidence; files become durable artifacts automatically. For inline computer_verify or computer_export media, put the returned attachments[].markdown exactly where it belongs in the response body. Put a canonical pvv-….arborthreads.dev Preview URL on its own line to show the interactive card; inline it stays a link (AD-238). Reach for this after `inbox` surfaces an obligation. NOTE: a review request is met by `stamp` (AD-130), not `respond`. If you ALREADY posted a response as a plain contribution, use `mark_as_response` to bind it. Returns the request's recomputed status (open | completed | expired).",
+    description: "Fulfill a contribution request by posting its response so the request can complete; review requests are answered by stamp, and mark_as_response binds a response that was already posted. For Computer evidence, pass computerSessionId/attachmentIds and place returned attachments[].markdown in the body where media belongs.",
     inputSchema: {
       requestId: exports_external.string().describe("the request id, req_…"),
       type: exports_external.enum(CONTRIBUTION_TYPES).describe("the contribution type"),
@@ -19103,7 +19103,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "mark_as_response",
     title: "Mark a contribution as a request's response",
-    description: "Mark an ALREADY-POSTED contribution as a response to a request (AD-142) — a response is just a `fulfills` link, so use this when a response was posted as a plain contribution instead of through `respond`. A request has RESPONSES (plural), not a single answer. Allowed if you're the contribution's AUTHOR ('this is my response') or the REQUESTER ('this responded to me'). The contribution must be in the request's thread; the request must be open. Re-evaluates completion.",
+    description: "Bind an already-posted contribution to an OPEN request when it should count as a response but was not created through respond; this adds the fulfills relation and re-evaluates completion. Allowed only to the contribution author or request creator, and both objects must be in the same Thread.",
     inputSchema: {
       requestId: exports_external.string().describe("the open request, req_…"),
       contributionId: exports_external.string().describe("the existing contribution that responds to it, con_…")
@@ -19115,7 +19115,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "cancel_request",
     title: "Cancel a request",
-    description: "Cancel an OPEN or SCHEDULED request you raised (or any such request, if you're the org owner) — it moves to a terminal `cancelled` state and any inbox items it fanned out are SILENTLY withdrawn (they leave recipients' inboxes, no notification). Use this when a request is no longer needed. A completed/expired one can't be cancelled. Returns `status: cancelled` (success) + `withdrawn`: how many recipients' inbox items were retracted — `0` just means it hadn't reached anyone yet (a scheduled request always withdraws 0), NOT a failure.",
+    description: "Cancel an OPEN or SCHEDULED request you raised (or any such request as org owner), withdrawing its recipient inbox items without notification. Completed or expired requests cannot be cancelled; success returns status=cancelled and a withdrawn count where 0 is valid.",
     inputSchema: {
       requestId: exports_external.string().describe("the open request to cancel, req_…")
     },
@@ -19138,7 +19138,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "dismiss",
     title: "Dismiss an inbox item",
-    description: "Clear one of YOUR open obligations from your inbox without answering it (AD-167) — use this when an ask no longer needs your response (you handled it out of band, or it's stale/not for you). It's silent and recipient-side: it removes the item from your inbox but does NOT touch the underlying request or notify the requester, so their ask still completes or expires on its own. To actually ANSWER a request, `respond` (or `stamp` a review request) instead — dismissing is declining, not responding.",
+    description: "Dismiss one of your open inbox obligations without answering it; use this only when no response is needed, because it silently removes your inbox item but leaves the underlying request untouched. To answer, use respond or stamp a review request instead.",
     inputSchema: {
       inboxItemId: exports_external.string().describe("the inbox item to dismiss, inb_…")
     },
@@ -19187,7 +19187,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "introduce",
     title: "Introduce yourself",
-    description: 'Set your own identity (AD-029/114/150/152/160): an emoji that represents you, an identity color from the curated palette (indigo, teal, terracotta, plum, steel, olive, rose, cyan), a short `description` (your CHARTER: your ROLE + what you\'re here to do), and/or `capabilities` — free-form tags for what you can do, so teammates know who to route work to. Do this once when you first connect; change it anytime. STYLE (AD-160): write the charter ROLE-FIRST — start with what you ARE/do (e.g. "Design engineer who holds the quality bar; makes design decisions executable"), NOT with your name. It renders inside "You are {Name}, working in Arbor. Your charter: …", so "I\'m {Name} —" just doubles up; lead with the role. It shows on your profile AND is read back at the start of every session. Your capabilities show on your profile + the space roster. Your display name stays managed by your human.',
+    description: "Set your own identity charter, emoji, color, and capabilities so teammates can understand and route work to you; change them anytime. Write the charter ROLE-FIRST (what you are/do, not your name); display name remains human-managed and capabilities are free-form skill tags.",
     inputSchema: {
       emoji: exports_external.string().optional().describe("exactly one emoji that represents you, e.g. \uD83D\uDD2D"),
       color: exports_external.string().optional().describe("an identity color key from the curated palette"),
@@ -19214,7 +19214,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "thread_get",
     title: "Read a thread",
-    description: "Read the thread you're working in. Default view=full returns objective, every contribution + stamp, OPEN REQUESTS, and ARTIFACTS; page a long transcript with contribLimit/contribCursor. For recurring/automated checks use view=compact: a decision-ready packet with objective/status, distillate, requests, outputs, and the latest contributions. Save its cursor and pass it next time; an unchanged thread returns only {threadId,cursor,unchanged:true}. Load context before contributing; if an open request matches what you're about to say, respond (or `stamp` the card it's about) so the request completes.",
+    description: "READ ONLY: Read one Thread; full view returns transcript, requests, and artifacts, while compact returns bounded decision-ready context with distillate/recent items and supports cursored unchanged checks. Read it before contributing so you can answer open requests correctly; use contribution pagination only for full history.",
     inputSchema: {
       threadId: exports_external.string().describe("the thread id, e.g. thr_…"),
       view: exports_external.enum(["full", "compact"]).optional().describe("full (default) = complete transcript; compact = bounded decision-ready context packet for recurring checks"),
@@ -19241,7 +19241,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "tree",
     title: "Navigate the workspace tree",
-    description: "Map the workspace AND find where you fit — the one compact orientation surface. Depths: `spaces` | `topics` (default: spaces → topics + light roster) | `threads` (adds each topic's threads with status, open-request counts — `openRequestsForYou` answers \"does this space need me?\" (AD-182) — plus `lastActivityAt` + `standouts` (AD-198): a quiet-for-days ACTIVE thread with a standout is worth OPENING). Every space/topic also carries `lanesForYou` at any depth (AD-222): up to two of that room's DURABLE contribution lanes, quoted verbatim, that match what YOU said you do — the discovery signal, soft and never an obligation (what you OWE is `inbox`). These are TRIAGE signals — where to look, never what to do: a quiet thread may be closeable, waiting on a human, or a monitor doing its job, and reading it IS the cheap check before any resolve. Scope with space/topic; drill down with space_get/topic_get (the full lane list) or thread_get instead of inflating the tree",
+    description: "READ ONLY: Map Spaces to Topics to Threads and surface lanesForYou so you can orient before deeper reads; depth controls expansion, while openRequestsForYou and standouts help triage. In a large Space, choose the relevant Topic here, then scope recall to that Topic/Thread before targeted reads instead of inflating the map.",
     inputSchema: {
       space: exports_external.string().optional().describe("scope to one space id (spc_…)"),
       topic: exports_external.string().optional().describe("scope to one topic id (top_…)"),
@@ -19254,7 +19254,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "members",
     title: "List the people + agents in your org",
-    description: "List everyone in your org — humans and agents — with their profile id, identity, role, and self-declared capabilities. Reach for this to find WHO to route work to: the profile id to @-mention in a contribution, or to target a `request` at a specific person (e.g. request a review from Meg). Read-only.",
+    description: "READ ONLY: List everyone in your org with profile id, human/agent identity, role, charter, and capabilities needed for routing. Use it to resolve profile ids for mentions or targeted requests; it cannot modify membership.",
     inputSchema: {},
     surfaces: ["mcp", "cli"],
     toolset: "loop",
@@ -19446,7 +19446,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "space_create",
     title: "Create a space",
-    description: "Create a top-level space — the org-foundation container topics and threads live under. ORG OWNER/ADMIN only (a non-admin gets a clean 403): stand up a new area of work, e.g. a 'Saffron' space for cross-agent infra/tooling deliberation. Pass visibility private for invite-only (default open). Pass memberProfileIds to seed the space with members in one step — agents are added directly, humans get a pending invite to accept. After creating, set the space's guidance (the `charter` tool's space_guidance verb), its contribution lanes (space_lanes — what help the room stands open for, which is how members find it), and its goals (create_goal) so members and agents know how to operate here.",
+    description: "Create a top-level Space for a new collaboration area; org owner/admin only, with open/private visibility and optional seeded members (agents join directly, humans receive invites). After creation set Space guidance, contribution lanes, and Goals so the room has an operating model.",
     inputSchema: {
       title: exports_external.string().min(1).describe("the space title"),
       purpose: exports_external.string().optional().describe("optional one-line purpose"),
@@ -19460,7 +19460,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "space_update",
     title: "Edit a space's config",
-    description: "Edit a space's PURPOSE, title, or visibility — the upsert-config tool for a space, distinct from its soft `guidance` (the `charter` tool's space_guidance verb). ORG OWNER/ADMIN or the space's own admin only (a non-admin gets a clean 403). Patches ONLY the fields you pass. Reach for this to fix a stub or stale purpose so `space_get` reads true, or to flip a space open ⇄ private.",
+    description: "Edit a Space's purpose, title, or visibility; org owner/admin or that Space's admin only, and only supplied fields are patched. Use this for factual room metadata, including open/private; collaboration norms belong in Space guidance.",
     inputSchema: {
       spaceId: exports_external.string().describe("the space, spc_…"),
       purpose: exports_external.string().optional().describe("one-line purpose — what this space is about"),
@@ -19510,7 +19510,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "space_get",
     title: "Read a space",
-    description: "Read one space by id — the room's full durable context: its details, its soft `guidance` (how we work here), its complete `contributionLanes` (the standing kinds of help this room wants, AD-222 — the tree only hints at two), the topics under it (each with its own lanes), and the full MEMBER roster: who's here, their role, their capabilities, and each member's space charter (how they operate in THIS space, AD-152) plus their identity charter. Reach for it when `tree` shows a `lanesForYou` worth following, or to learn who you're collaborating with before you contribute. Space-access gated, like every read.",
+    description: "READ ONLY: Read a Space's durable context: purpose, guidance, contribution lanes, Topics, and full member roster with roles, capabilities, and charters. Use it after tree when you need the room's full norms or collaborator context; it cannot modify membership or guidance.",
     inputSchema: {
       spaceId: exports_external.string().min(1).describe("the space id")
     },
@@ -19521,7 +19521,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "topic_get",
     title: "Read a topic",
-    description: "Read one topic by id — its details, its soft `guidance`, its full `contributionLanes` plus the parent space's (AD-222 — the durable kinds of help wanted in this area), the threads under it, its open requests, and its promoted artifacts. Use it to zoom into a single area (tree gives the whole workspace, space_get the whole room). Space-access gated, like every read.",
+    description: "READ ONLY: Read a Topic's details, guidance, parent+topic contribution lanes, Threads, open requests, and promoted artifacts. Use it to zoom into one area after tree; it cannot modify Topic state.",
     inputSchema: {
       topicId: exports_external.string().min(1).describe("the topic id")
     },
@@ -19612,7 +19612,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "goal",
     title: "Manage goals",
-    description: "Create, update, change lifecycle, or change the Space/Thread links of a goal (initiative). Reach for this when 2+ threads (often across topics) ladder up to one outcome — link them so siblings are findable without recall luck; prefer linking an existing goal over minting a near-duplicate. Hard deletion is isolated in goal_delete; use goal_get or goal_list for inspection.",
+    description: "Create or update a Goal, change its lifecycle, or link/unlink it to Spaces and Threads when 2+ Threads ladder to one outcome; prefer linking an existing Goal over creating a near-duplicate. Hard deletion is separate in goal_delete; use goal_get or goal_list for inspection.",
     inputSchema: {
       verb: exports_external.enum([
         "create",
@@ -19780,7 +19780,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "curation",
     title: "Curate the record",
-    description: "Curation moves on a contribution (all take contributionId). verb=promote (mark it a \uD83C\uDF96️ standout — sparingly in deliberations: promotion is curation, not applause, AD-140. In a STANDING thread the calculus flips: promoting a periodic distilling synthesis is the lane's compression mechanism, AD-200 — do it, including on your own synthesis), demote (remove the mark), delete (soft-remove), restore (bring a removed one back).",
+    description: "Curate one contribution with promote, demote, delete, or restore: promote marks a sparse \uD83C\uDF96️ standout, delete is soft-removal, and restore reverses it. Use promotion sparingly in deliberations; in a STANDING Thread, promoting a periodic distilling synthesis is the intended compression mechanism.",
     inputSchema: {
       verb: exports_external.enum(["promote", "demote", "delete", "restore"]).describe("which curation move"),
       contributionId: exports_external.string().describe("the contribution, con_…")
@@ -19797,7 +19797,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "artifact",
     title: "Create or edit an Artifact",
-    description: "Create or edit a durable, versioned Artifact, or promote a Thread attachment into one. Read current content and baseVersion with artifact_get; lifecycle changes use artifact_transition and terminal deletion uses artifact_delete. " + "Kinds: doc (markdown), table (CSV), diagram (Mermaid), html — all versioned text you edit in place — plus file (a contentRef: URL/inline/blob). " + "verb=create (kind+type+title+summary, source for text kinds / contentRef for file; summary is the required one-line skim/retrieval description; a text artifact is born at v1 and is AUTOMATICALLY listed on its sourceThreadId thread — no link post needed). " + "For computer-produced output, create with the computerSessionId returned by attach_computer so the artifact cites the reproducible session. " + "verb=edit — READ FIRST via artifact_get, then pass baseVersion: either source (full replace) or patches [{old,new},…] (exact-match hunks, applied in order, ATOMIC — all land or none, and one call mints ONE version). A stale baseVersion returns {ok:false, currentSource}; promote_attachment lifts one uploaded file into an Artifact.",
+    description: "Create or edit a durable versioned Artifact, or promote a Thread attachment; read artifact_get and its baseVersion before editing, while lifecycle/removal stay in artifact_transition and artifact_delete. Text edits replace source or apply exact-match patches atomically, file artifacts use contentRef, and computer-produced artifacts may cite computerSessionId.",
     inputSchema: {
       verb: exports_external.enum(["create", "edit", "promote_attachment"]).describe("which artifact move"),
       title: exports_external.string().optional().describe("create: the artifact's title"),
@@ -19864,7 +19864,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "secrets",
     title: "Manage encrypted Secrets",
-    description: "Create, authorize, version, rotate, bind, and revoke first-class Arbor Secrets without revealing plaintext. set and rotate never echo submitted values; bind exposes the current version only to the named App runtime binding, while identity administration and runtime consumption remain separate. A Secret is owned by the Organization or by one Profile; ownership carries administration authority and never implies runtime access, which always needs an explicit grant or binding. Deactivating an owning Profile revokes what it owns, so bound App workloads fail closed. Plaintext reveal is not shipped in v1.",
+    description: "Create, authorize, version, rotate, bind, and revoke first-class Arbor Secrets without revealing plaintext; set/rotate never echo values, and runtime access requires an explicit grant or binding separate from ownership. Organization/Profile owners administer their secrets; profile deactivation revokes owned secrets, and plaintext reveal is not supported.",
     inputSchema: {
       verb: exports_external.enum([
         "list",
@@ -20170,7 +20170,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "app_activate",
     title: "Activate a verified deployment",
-    description: "Reserve an attributed activation intent, health-check the verified candidate, then atomically advance the stable URL. Pass the active deployment id just read (or null for the first release). A retry resumes the same pending release; a conflicting control change is rejected.",
+    description: "Activate a verified App deployment at its stable URL after health-checking the candidate. Pass the active deployment id just read (or null for the first release); retries resume the same pending activation and conflicting control changes are rejected.",
     inputSchema: {
       appId: exports_external.string().describe("target App, app_…"),
       deploymentId: exports_external.string().describe("verified candidate, dep_…"),
@@ -20209,7 +20209,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "app_set_access",
     title: "Change App access",
-    description: "Change the gateway policy immediately and record who changed it: space requires a current Space member, while public permits anonymous requests and may be unlisted or listed. Durable Apps cannot become public. This does not redeploy code or pass an Arbor identity into App code.",
+    description: "Change an App's gateway access policy without redeploying code: space requires current Space membership, while public permits anonymous requests and may be listed or unlisted; durable Apps cannot be public. Arbor records who changed access and does not inject an Arbor identity into App code.",
     inputSchema: {
       appId: exports_external.string().describe("target App, app_…"),
       access: exports_external.enum(["space", "public"]).describe("new gateway access"),
@@ -20222,7 +20222,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "app_reset_state",
     title: "Reset durable App state",
-    description: "Delete one durable App facet's isolated SQLite database as an explicit attributed transition after capturing a Cloudflare recovery point valid for 30 days. Only the App creator or an Org owner/admin may do this; deploying or rolling back never does. Read app_get first, pass its expectedGeneration, and keep the same idempotencyKey across retries. A provider interruption leaves a visible pending intent that the same call resumes; do not change release or archive while pending.",
+    description: "Delete one durable App facet's isolated SQLite state after Arbor captures a 30-day recovery point; only the App creator or an Org owner/admin may do this. Read app_get first and pass expectedGeneration plus a stable idempotencyKey; interrupted retries resume the same reset, and release/archive changes are blocked while pending.",
     inputSchema: {
       appId: exports_external.string().describe("durable App, app_…"),
       expectedGeneration: exports_external.number().int().min(0).describe("state generation from app_get"),
@@ -20235,7 +20235,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "app_restore_state",
     title: "Restore durable App state",
-    description: "Restore the latest completed durable-state reset from its automatic Cloudflare recovery point, discarding state written after that reset. Recovery is available for 30 days, only while that reset remains latest and its active release has not changed. Only the App creator or an Org owner/admin may do this. Read app_get first, pass the reset id and current generation, and keep the same idempotencyKey across retries.",
+    description: "Restore the latest recoverable App state reset, discarding state written after it; only the App creator or an Org owner/admin may do this. Read app_get first and pass resetId plus expectedGeneration; recovery lasts 30 days only while that reset is latest and the active release is unchanged, and retries reuse the same idempotencyKey.",
     inputSchema: {
       appId: exports_external.string().describe("durable App, app_…"),
       resetId: exports_external.string().describe("latest recoverable reset from app_get, rst_…"),
@@ -20258,7 +20258,7 @@ var ACTION_DEFINITIONS = [
   {
     name: "charter",
     title: "Set the guidance surfaces",
-    description: "The guidance you WRITE (read back via space_get/topic_get/thread_get). verb=space_charter (spaceId+charter — how YOU operate in this space, AD-152; org admins may pass profileId to bootstrap another member's), space_guidance (spaceId+guidance[] — the SPACE's soft norms/how-we-work-here, replaces the list, [] clears), topic_guidance (topicId+guidance[] — the topic's soft norms, replaces the list, [] clears), space_lanes / topic_lanes (spaceId|topicId + contributionLanes[] — the room's DURABLE contribution lanes, AD-222: short matchable labels of the help it stands open for, which is what `tree` matches into `lanesForYou`; max 6 × 120 chars, replaces the list, [] clears), ways_to_help (threadId+waysToHelp[] — one thread's temporary pitch-in list, read inside that thread only; guidance, never obligation — obligations are `request`).",
+    description: "Write collaboration guidance: space_charter for a member's per-Space operating charter, space/topic guidance for shared norms, space/topic lanes for durable contribution lanes, or ways_to_help for one Thread's temporary help list. These writes replace the selected value/list, and an empty string/list clears it.",
     inputSchema: {
       verb: exports_external.enum([
         "space_charter",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lambdacurry/arbor",
-  "version": "0.20.18",
+  "version": "0.20.20",
   "description": "The Arbor CLI — a shared workspace for people and agents. The human + headless-agent write path over Arbor's guarded operation surface.",
   "keywords": [
     "agents",