@sellable/mcp 0.1.311 → 0.1.313

package/README.md

@@ -66,7 +66,7 @@ surfaces that are not posts or campaigns.
 For Claude Code and Codex CLI:
 
 ```bash
-npx -y @sellable/install@latest --host all --token skt_live_your_token_here --workspace-id your_workspace_id
+curl -fsSL "https://app.sellable.dev/api/v2/cli/install" | sh
 ```
 
 Agent-readable install instructions are available at:
@@ -78,8 +78,13 @@ https://app.sellable.dev/agent-install.txt
 The reviewable curl installer is:
 
 ```bash
-curl -fsSL "https://app.sellable.dev/api/v2/cli/install" -o /tmp/sellable-install.sh
-sh /tmp/sellable-install.sh
+curl -fsSL "https://app.sellable.dev/api/v2/cli/install" | sh
+```
+
+Windows users can use the native PowerShell installer:
+
+```powershell
+iwr "https://app.sellable.dev/api/v2/cli/install.ps1" | iex
 ```
 
 Phase 112 v1 uses package stdio MCP through `@sellable/mcp`. Hosted HTTP
@@ -88,7 +93,15 @@ MCP is a future/advanced mode until the hosted endpoint exists.
 Package-mode installs launch `@sellable/mcp@latest`, so each fresh host MCP
 start resolves the newest stable npm release. The MCP server also runs a cached
 startup/auth update check and tells the agent to run
-`npx -y @sellable/install@latest` when the installed runtime is behind npm.
+`curl -fsSL "https://app.sellable.dev/api/v2/cli/install" | sh` when the
+installed runtime is behind npm.
+
+For CI/scripted installs, set `SELLABLE_TOKEN` and `SELLABLE_WORKSPACE_ID`
+before running the curl installer. The public path remains:
+
+```bash
+curl -fsSL "https://app.sellable.dev/api/v2/cli/install" | sh
+```
 
 ### Publishing `@sellable/mcp`
 
@@ -110,12 +123,16 @@ also supported.
 2. Click "Generate Token"
 3. Copy the token (starts with `skt_live_`)
 
-### 3. Claude Setup
+### 3. Claude Setup Fallback
+
+The public installer should handle Claude setup. Use this manual path only for
+troubleshooting when Node/npm/npx are already known-good and the installer route
+is unavailable.
 
-Install MCP server:
+Install MCP server manually:
 
 ```bash
-claude mcp add --transport stdio sellable -- npx -y @sellable/mcp@latest
+claude mcp add --transport stdio sellable -- npm exec --yes --package @sellable/mcp@latest -- sellable-mcp
 ```
 
 Create auth config at `~/.sellable/config.json`:
@@ -135,15 +152,15 @@ The token is provided when you generate it. Use `list_workspaces` +
 For customer/package installs, use the public installer:
 
 ```bash
-npx -y @sellable/install@latest --host codex --token skt_live_your_token_here --workspace-id your_workspace_id
+curl -fsSL "https://app.sellable.dev/api/v2/cli/install" | sh
 ```
 
 If you already have `~/.sellable/config.json`, rerun/verify without rewriting
 auth:
 
 ```bash
-npx -y @sellable/install@latest --host codex
-sellable --verify-only --host codex
+curl -fsSL "https://app.sellable.dev/api/v2/cli/install" | sh
+sellable --verify-only --host codex --json
 ```
 
 The installer does the full local setup:

package/dist/tools/bootstrap.js

@@ -12,7 +12,7 @@ function toGuidance(check, message) {
         return "Run get_auth_status and fix auth/workspace setup before continuing.";
     }
     if (check === "update") {
-        return "Run npx -y @sellable/install@latest, then restart Claude Code/Codex so Sellable reloads.";
+        return 'Run curl -fsSL "https://app.sellable.dev/api/v2/cli/install" | sh, then restart Claude Code/Codex so Sellable reloads.';
     }
     if (check === "framework") {
         return "Fix framework files and rerun bootstrap_create_campaign.";

package/dist/tools/campaign-message-preparation.d.ts

@@ -11,6 +11,7 @@ type StartPrepareMessagesInput = PrepareMessagesBaseInput & {
     batchSize?: number;
     approvalMode?: ApprovalMode;
     autoContinue?: boolean;
+    disableLowPassRateStop?: boolean;
 };
 type StatusPrepareMessagesInput = PrepareMessagesBaseInput;
 type CancelPrepareMessagesInput = PrepareMessagesBaseInput;
@@ -52,6 +53,10 @@ export declare const campaignMessagePreparationToolDefinitions: ({
             autoContinue: {
                 type: string;
             };
+            disableLowPassRateStop: {
+                type: string;
+                description: string;
+            };
             jobId?: undefined;
         };
         required: string[];
@@ -78,6 +83,7 @@ export declare const campaignMessagePreparationToolDefinitions: ({
             batchSize?: undefined;
             approvalMode?: undefined;
             autoContinue?: undefined;
+            disableLowPassRateStop?: undefined;
         };
         additionalProperties: boolean;
         required?: undefined;

package/dist/tools/campaign-message-preparation.js

@@ -6,7 +6,7 @@ async function postPrepareMessages(body) {
 export const campaignMessagePreparationToolDefinitions = [
     {
         name: "start_campaign_message_preparation",
-        description: 'Start a bounded campaign message preparation job for a CampaignOffer campaignId. Use this after lead/message approval only when the user asks for a target like "prepare 100 messages" or "schedule 250 sends"; default approvalMode is mark_ready and it never starts the campaign. Omit maxRowsToCheck and batchSize for the enforced adaptive default: first calibrate on at least 100 rows, estimate the row budget from observed rubric/pass yield, cap rows at 2500, then use batches up to 250 once the sample is strong enough.',
+        description: 'Start a bounded campaign message preparation job for a CampaignOffer campaignId. Use this after lead/message approval when the user asks to "fill up", "load", "prepare", or "schedule" sends for attached senders. It never launches the campaign. The job queues pending Enrich Prospect cells first, lets ICP/rubric and Generate Message cascade, then marks ready or approves only the bounded cohort. Omit maxRowsToCheck and batchSize for the adaptive default: calibrate on at least 100 actually-enriched rows, estimate the row budget from observed rubric/pass yield, cap rows at 2500, then use batches up to 250 once the sample is strong enough. Do not interpret checkedRows as enriched rows; use progress.enrichedRows, needsEnrichRows, activeCellCount, preparedMessages, and stopReason.',
         inputSchema: {
             type: "object",
             properties: {
@@ -34,6 +34,10 @@ export const campaignMessagePreparationToolDefinitions = [
                     description: "Defaults to mark_ready. Use approve only when the user explicitly asks to flip Approved cells.",
                 },
                 autoContinue: { type: "boolean" },
+                disableLowPassRateStop: {
+                    type: "boolean",
+                    description: "Operator override for explicit fill/exhaust requests. When true, keep processing eligible rows even if the enriched sample pass rate is low. Still bounded by targetPreparedMessages and maxRowsToCheck.",
+                },
             },
             required: ["campaignId"],
             additionalProperties: false,
@@ -76,6 +80,7 @@ export function startPrepareCampaignMessages(input) {
         batchSize: input.batchSize,
         approvalMode: input.approvalMode,
         autoContinue: input.autoContinue,
+        disableLowPassRateStop: input.disableLowPassRateStop,
     });
 }
 export function getPrepareCampaignMessagesStatus(input) {

package/dist/tools/model-quality.d.ts

@@ -14,6 +14,7 @@ export type CampaignModelQualityResult = {
     recommendedReasoningEffort: string;
     minimumSummary: string;
     confirmationRequired: boolean;
+    metadataStale: boolean;
     message: string;
 };
 export type CampaignModelQualityHostConfig = {
@@ -33,6 +34,7 @@ export type CampaignModelQualityConfig = {
     };
     warningCopy: {
         ok: string;
+        staleCodexMetadata: string;
         unknownSettings: string;
         belowMinimum: string;
     };

package/dist/tools/model-quality.js

@@ -37,6 +37,7 @@ const DEFAULT_MODEL_QUALITY_CONFIG = {
     },
     warningCopy: {
         ok: "Campaign model settings meet the configured minimum: {minimumSummary}.",
+        staleCodexMetadata: "Codex host metadata appears stale: it can report GPT-5/default or GPT-5/high even when the UI is set to GPT 5.5 Extra High. Treating this as acceptable; continue without asking the user to switch.",
         unknownSettings: "Model settings were not provided by the host. Best campaigns need at least {minimumSummary}. Confirm the user is on one of those settings, or ask them to switch before continuing.",
         belowMinimum: "Best campaigns need at least {minimumSummary}. Current settings look below that: {currentSettings}. Please switch before continuing, or explicitly say to continue anyway.",
     },
@@ -146,6 +147,22 @@ function findAcceptedHostConfig(host, model, config) {
         : [[host, config.hosts[host]]];
     return candidates.find(([, hostConfig]) => modelMeetsMinimum(model, hostConfig));
 }
+function looksLikeCodexStaleMetadata(host, model, reasoningEffort, config) {
+    if (host !== "codex")
+        return false;
+    if (!model || !reasoningEffort)
+        return false;
+    const normalizedModel = normalize(model).replace(/[_-]+/g, " ");
+    const normalizedReasoning = normalizeReasoning(reasoningEffort);
+    if (!normalizedModel.includes("gpt"))
+        return false;
+    const versions = extractModelVersions(normalizedModel);
+    const looksLikeBaseGpt5 = versions.some((version) => version === "5" || version === "5.0");
+    if (!looksLikeBaseGpt5)
+        return false;
+    return (["default", "auto", "standard"].includes(normalizedReasoning) ||
+        acceptsReasoning(reasoningEffort, config.hosts.codex));
+}
 export function evaluateCampaignModelQuality(input = {}) {
     const config = getCampaignModelQualityConfig();
     const host = normalizeHost([input.host, input.model].filter(Boolean).join(" "));
@@ -166,6 +183,7 @@ export function evaluateCampaignModelQuality(input = {}) {
             recommendedReasoningEffort,
             minimumSummary,
             confirmationRequired: true,
+            metadataStale: false,
             message: formatCopy(config.warningCopy.unknownSettings, {
                 minimumSummary,
                 currentSettings: "unknown",
@@ -185,12 +203,30 @@ export function evaluateCampaignModelQuality(input = {}) {
                 recommendedReasoningEffort,
             minimumSummary,
             confirmationRequired: false,
+            metadataStale: false,
             message: formatCopy(config.warningCopy.ok, {
                 minimumSummary,
                 currentSettings: "current settings",
             }),
         };
     }
+    if (looksLikeCodexStaleMetadata(host, model, reasoningEffort, config)) {
+        return {
+            status: "ok",
+            host,
+            model,
+            reasoningEffort,
+            recommendedModel,
+            recommendedReasoningEffort,
+            minimumSummary,
+            confirmationRequired: false,
+            metadataStale: true,
+            message: formatCopy(config.warningCopy.staleCodexMetadata, {
+                minimumSummary,
+                currentSettings: "stale Codex host metadata",
+            }),
+        };
+    }
     const currentSettings = [
         model ? `model "${model}"` : "unknown model",
         reasoningEffort ? `reasoning "${reasoningEffort}"` : "unknown reasoning",
@@ -204,6 +240,7 @@ export function evaluateCampaignModelQuality(input = {}) {
         recommendedReasoningEffort,
         minimumSummary,
         confirmationRequired: true,
+        metadataStale: false,
         message: formatCopy(config.warningCopy.belowMinimum, {
             minimumSummary,
             currentSettings,

package/dist/tools/prompts.js

@@ -367,7 +367,7 @@ export function getPostFindLeadsScoutRegistry() {
             codex: 'After confirm_lead_list copies source rows and the initial campaign-table execution slice exists, ask the filter-choice question immediately. Do not spawn anything before that question. After the answer, launch only Message Drafting. The filter-choice answer is the post-import user gate for this single worker; do not ask another question about starting it in step-wise or YOLO mode. The registry lookup is not a launch: after get_post_find_leads_scout_registry, immediately invoke Task/spawn_agent or the host background-agent mechanism before loading filter-leads.md, before saving rubrics, and before treating skip-filters as ready for message review. Both choices route through this kickoff; do not let filters_skipped jump straight from filter-choice to message-generation. If filters are chosen, the parent stays on Filter Rules and drafts/saves rubrics with MCP tools while Message Drafting runs in the background. If filters are skipped, move to Messages/message review only after Message Drafting has started or is ready; update_campaign(currentStep=messages) is not proof of launch. If the named Message Drafting custom agent is unavailable, spawn a generic gpt-5.5 xhigh Message Drafting background agent with the same lean campaign/table basis. When the background worker starts, persist workerDetails.messageDraftBuilder with statusSource "branch", status "branch-running", runId, startedAt, updatedAt, basisToken when known, and basis containing campaignId, selectedLeadListId, workflowTableId, filterChoice, and reviewBatchRowHash or reviewBatchRowIds; workerStatuses.messageDraftBuilder may be "running" as a simple badge only. Never put rich proof under workerStatuses and never use workerStatuses.messageDrafting. If no background-agent tool is callable, start the same full message branch inline before filter drafting or before skip-filter message review, record workerDetails.messageDraftBuilder with statusSource "parent-thread-fallback" and status "fallback-active", and require the same live context, prompt, assets, and validation gate before message review; do not wait until filters are saved and then call the registry.',
             claude: "After confirm_lead_list copies source rows and the initial campaign-table execution slice exists, ask the filter-choice question immediately. Do not invoke any Task/Agent before that question. After the answer, invoke only Message Drafting. If filters are chosen, parent drafts/saves rubrics with MCP tools while Message Drafting runs, asks filter approval, then joins Message Drafting. If filters are skipped, invoke only Message Drafting and move to Messages/message review.",
             parentThreadRule: 'Named agents are optional acceleration, but message drafting is not optional. The only normal background worker is Message Drafting. The filter-choice answer is the campaign-scoped go-ahead for this single post-import worker; do not ask another question to start it in step-wise or YOLO mode. If a named agent is unavailable, use a generic gpt-5.5 xhigh Message Drafting background agent. source work and filter work stay in the parent thread with MCP tools. If post-find-leads-message-scout is available, run it as the background Message Draft Builder after the filter-choice answer. The registry lookup is not a launch: get_post_find_leads_scout_registry only identifies the worker, and Message Drafting counts as started only after Task/spawn_agent or the host background-agent tool is invoked, or after the parent begins the same full message branch inline because no background-agent tool is callable. This launch must happen before loading filter-leads.md, save_rubrics, filter approval, or skip-filter message review; currentStep=messages is not proof of launch. If post-find-leads-message-scout is absent, do not customer-surface install status. Do not silently treat message drafting as started; the main thread must either launch the background worker or execute the same message branch from CampaignOffer state, selected source state, workflowTableId, and initial campaign-table execution slice rows. For a spawned worker, record workerDetails.messageDraftBuilder with statusSource branch / status branch-running, runId, startedAt, updatedAt, and basis containing campaignId, selectedLeadListId, workflowTableId, filterChoice, and reviewBatchRowHash or reviewBatchRowIds. workerStatuses.messageDraftBuilder is optional simple badge text only ("running", "ready", "blocked", "idle"); never put runId/statusSource/basis under workerStatuses and never use workerStatuses.messageDrafting. If no background-agent tool is callable, start that same full message branch inline before filter drafting or before skip-filter message review, record workerDetails.messageDraftBuilder with statusSource parent-thread-fallback / status fallback-active then ready, and require the same live context, prompt, assets, and validation gate before message review; do not report that as a background worker failure. If neither branch nor inline fallback can run, return blocked/retry-needed; do not wait until filters are saved and then call the registry. The Message Drafting handoff must be lean. Do not paste copied row counts, brief hashes, review-batch hashes, full reviewBatchRowIds, broad row data, or local debug artifacts into the spawn prompt. Local markdown/json files are not normal-path inputs. The filter-choice question is the first post-import user gate; do not load post-lead registries or filter references before it. Message drafting starts after the filter-choice answer, must load get_subskill_prompt({ subskillName: "generate-messages" }), and must load every required message asset named by generate-messages Mode 0 through get_subskill_asset before drafting. Reference Asset Loading means loading the required pre-draft reference pack before drafting; return blocked/retry-needed if required assets cannot be loaded; load ai-tells.md because it is never optional. The branch or parent-thread fallback loads the full generate-messages prompt and every referenced asset through get_subskill_asset. After generating/revising the candidate and before returning ready, must load get_subskill_prompt({ subskillName: "create-campaign-v2-validation" }) as the final internal validation gate, must read live campaign table state through scoped MCP/product tools, and must reject mismatched selectedLeadListId/workflowTableId/campaign/workspace input. Do not block when filters were chosen but leadScoringRubrics are not yet visible in the branch read; the parent owns save_rubrics and filter approval in parallel, so Message Drafting should return status ready with basisStatus usable_initial when campaign/list/table identity and the non-empty execution slice match. Do not use any alternate, local-artifact, or examples-only message prompt. User copy feedback, message QA, or rewrite requests before approve-message must be routed back to Message Drafting with the current recommendation, lean campaign/table basis, and latest user text; the parent must not rewrite or QA the template from memory and must not call update_campaign_brief before approve-message. The worker validates internally and returns only templateRecommendation, tokenFillRules, renderedGoodSample, status, approveOrReviseRecommendation, validationStatus, outputAt, outputHash, and blocked/retry detail. Do not render renderedFallbackSample, risk notes, or a qaReceipt on the normal happy path. On the filter path, save_rubrics keeps the browser on Filter Rules after save_rubrics so the user can approve the saved criteria; after saved-filter approval, move to Filter Leads with currentStep=apply-icp-rubric whether Message Drafting is ready or still running. Wait there for message approval. Enrichment, filtering, Generate Message cells, sender setup, sequence attach, and launch wait for template approval on the Use Template path. On the skip path, move to Messages/message review after Message Drafting has started or is ready and wait for message approval before enrichment or Settings. Do not render message review from checklist or shortcut instructions; message review requires a messageDraftRecommendation whose basis proves the generate-messages prompt, required message assets, and validation gate ran for the current campaign/table execution slice. Do not automatically rerun Message Drafting after filters/enrichment finish; show the initial draft by default and offer an enriched rewrite only with explicit user opt-in. Handoff and recommendation output are Markdown with labeled fields, not raw JSON.',
-            prepareMessagesRule: 'Default create-campaign stays on the existing reviewBatchLimit:15 first campaign-table execution slice. Only call start_campaign_message_preparation when the user explicitly asks for more prepared messages or a send count. For "prepare/generate X messages", set targetPreparedMessages:X, omit maxRowsToCheck so the backend calibrates on at least 100 rows, estimates the row budget from observed rubric/pass yield, caps maxRowsToCheck at 2500, and use approvalMode:mark_ready. After the calibration sample settles, the backend adapts later batches up to 250 rows while recalculating yield. Poll get_campaign_message_preparation_status and summarize preparation-job status: checked rows, passed/prepared/approved count, target, estimated row budget remaining, and stop reason. For "approve X messages", use approvalMode:approve but still do not launch. For "schedule X sends", use approvalMode:approve to approve exactly the bounded X-message cohort during preparation, then continue through sender, sequence, and final launch greenlight; final launch must verify that bounded cohort and must not broad approve-all. campaignId is CampaignOffer.id. If the user asks to stop preparation, the target is wrong, or status shows the wrong campaign/table, call cancel_campaign_message_preparation; otherwise do not cancel a healthy prepare run. cancel_campaign_message_preparation cancels the same pending workflow-table cells as the UI Cancel Pending Cells action. Low-level selectors are diagnostics and recovery only for this lane. start_campaign remains forbidden until final launch greenlight.',
+            prepareMessagesRule: 'Default create-campaign stays on the existing reviewBatchLimit:15 first campaign-table execution slice. Only call start_campaign_message_preparation when the user explicitly asks for more prepared messages, a send count, or to fill up/load sends for senders. Treat "fill up/load sends" as capacity-fill preparation: calculate the bounded target from sender capacity when needed, then let the job queue pending Enrich Prospect cells first, wait for ICP/rubric and Generate Message to cascade, and mark ready or approve only the target cohort. Do not interpret checkedRows as enriched rows; it is only the table cursor. Poll get_campaign_message_preparation_status and summarize enrichedRows, needsEnrichRows, activeCellCount, passed/prepared/approved count, target, estimated row budget remaining, and stopReason. For "prepare/generate X messages", set targetPreparedMessages:X, omit maxRowsToCheck so the backend calibrates on at least 100 actually-enriched rows, estimates the row budget from observed rubric/pass yield, caps maxRowsToCheck at 2500, and use approvalMode:mark_ready. After the calibration sample settles, the backend adapts later batches up to 250 rows while recalculating yield. For "approve X messages", use approvalMode:approve but still do not launch. For "schedule X sends" or "fill sender sends", use approvalMode:approve to approve exactly the bounded X-message cohort during preparation, then continue through sender, sequence, and final launch greenlight; final launch must verify that bounded cohort and must not broad approve-all. campaignId is CampaignOffer.id. If the user asks to stop preparation, the target is wrong, or status shows the wrong campaign/table, call cancel_campaign_message_preparation; otherwise do not cancel a healthy prepare run. cancel_campaign_message_preparation cancels the same pending workflow-table cells as the UI Cancel Pending Cells action. Low-level selectors are diagnostics and recovery only for this lane. start_campaign remains forbidden until final launch greenlight.',
         },
     };
 }

package/dist/tools/registry.d.ts

@@ -508,6 +508,10 @@ export declare const allTools: ({
             autoContinue: {
                 type: string;
             };
+            disableLowPassRateStop: {
+                type: string;
+                description: string;
+            };
             jobId?: undefined;
         };
         required: string[];
@@ -534,6 +538,7 @@ export declare const allTools: ({
             batchSize?: undefined;
             approvalMode?: undefined;
             autoContinue?: undefined;
+            disableLowPassRateStop?: undefined;
         };
         additionalProperties: boolean;
         required?: undefined;

package/dist/update-check.js

@@ -159,7 +159,7 @@ function buildNotice(mcp, installer) {
     const installerVersion = installer.latestVersion
         ? ` The latest installer is ${installer.latestVersion}.`
         : "";
-    return `Sellable update available: ${versions}. Run: npx -y @sellable/install@latest.${installerVersion}`;
+    return `Sellable update available: ${versions}. Run: curl -fsSL "https://app.sellable.dev/api/v2/cli/install" | sh.${installerVersion}`;
 }
 export async function checkForUpdates(options) {
     const currentVersion = readCurrentVersion();
@@ -212,7 +212,9 @@ export async function checkForUpdates(options) {
         installer,
         updateAvailable: mcp.outdated,
         blocking: false,
-        guidance: notice ? "Run npx -y @sellable/install@latest" : null,
+        guidance: notice
+            ? 'Run curl -fsSL "https://app.sellable.dev/api/v2/cli/install" | sh'
+            : null,
         _userNotice: notice,
     };
     writeCache(status);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@sellable/mcp",
-  "version": "0.1.311",
+  "version": "0.1.313",
   "type": "module",
   "description": "Sellable MCP server for Claude Code and Codex campaign workflows",
   "main": "dist/index.js",

package/skills/create-campaign/SKILL.md

@@ -100,20 +100,25 @@ The default path stays the existing first campaign-table execution slice:
 review the normal `reviewBatchLimit:15`, approve reviewed draft rows, then move
 to Settings/sequence/final greenlight. Only call
 `start_campaign_message_preparation` when the user explicitly asks for more
-prepared messages or a send count. If the user says "prepare/generate X
+prepared messages, a send count, or language like "fill up/load sends for these
+senders." Treat those requests as capacity-fill preparation: calculate the
+bounded target from sender capacity when needed, then let the preparation job
+queue pending `Enrich Prospect` cells, wait for ICP/rubric and Generate Message
+cells to cascade, and mark ready or approve only the target cohort. Do not
+count `checkedRows` as enriched rows; it is only the table cursor. Use
+`progress.enrichedRows`, `progress.needsEnrichRows`, `activeCellCount`,
+`preparedMessages`, `approvedRows`, target, estimated row budget remaining, and
+`stopReason` to explain progress. If the user says "prepare/generate X
 messages", set `targetPreparedMessages:X`, omit `maxRowsToCheck`, and keep
-`approvalMode:"mark_ready"`. The backend calibrates on at least 100 rows,
-estimates the row budget from observed rubric/pass yield, caps
+`approvalMode:"mark_ready"`. The backend calibrates on at least 100 actually
+enriched rows, estimates the row budget from observed rubric/pass yield, caps
 `maxRowsToCheck` at 2500, then adapts later batches up to 250 rows while
-recalculating yield. Poll
-`get_campaign_message_preparation_status` for preparation-job status: checked
-rows, passed/prepared/approved count, target, estimated row budget remaining,
-and stop reason. If the user says "approve X messages", use
+recalculating yield. If the user says "approve X messages", use
 `approvalMode:"approve"` but still do not launch. If the user says "schedule X
-sends", use `approvalMode:"approve"` to approve exactly the bounded X-message
-cohort during preparation, then continue through sender, sequence, and final
-launch greenlight; the launch path must verify that bounded cohort and must not
-broad approve-all.
+sends" or asks to fill sender sends, use `approvalMode:"approve"` to approve
+exactly the bounded X-message cohort during preparation, then continue through
+sender, sequence, and final launch greenlight; the launch path must verify that
+bounded cohort and must not broad approve-all.
 When approving reviewed draft rows in the campaign table, resolve the actual
 visible `Approved` cells with `select_campaign_cells({ columnRole: "approved",
 rowSelector: { type: "rowIds", rowIds } })` and `update_cell` those returned
@@ -174,11 +179,17 @@ person/company this campaign is for, then I’ll turn that into a campaign brief
 before we move into lead sourcing.
 ```
 
-Exception: if `bootstrap_create_campaign.modelQuality.status === "warn"`,
-the first visible campaign message must be the model-quality warning from
+Exception: if `bootstrap_create_campaign.modelQuality.status === "warn"` and
+`bootstrap_create_campaign.modelQuality.metadataStale !== true`, the first
+visible campaign message must be the model-quality warning from
 `modelQuality.message`. Ask the user to switch to the configured minimum model
 or explicitly continue anyway before identity setup, research, lead filtering,
-message generation, or launch review.
+message generation, or launch review. If `metadataStale === true`, continue
+normally and do not ask the user to switch.
+
+If `bootstrap_create_campaign.modelQuality.metadataStale === true`, continue
+normally. Do not ask the user to switch models; this is an accepted Codex host
+metadata mismatch.
 
 If a linked/local skill file is stale or missing, silently use the installed
 `sellable@sellable` plugin copy. Do not tell the user about the stale link,
@@ -807,12 +818,15 @@ there.
 MCP tool access is required. First call `mcp__sellable__get_auth_status({})`
 directly. If that tool is unavailable, stop and say this is a Codex
 install/reload problem, not a campaign problem. Tell the user to
-run `npx -y @sellable/install@latest --host all` so the packaged MCP server,
-Codex Desktop plugin, and Sellable skill bundle are installed. If they want a
-CLI verification, tell them to run `sellable --verify-only --host all`. After
-that, they must fully quit and reopen Codex Desktop before starting a new
-thread. Do not use `scripts/mcp/sellable-tool-call.mjs`, `npm run`,
-`node`, or any local harness as a fallback for this interactive skill.
+run `curl -fsSL "https://app.sellable.dev/api/v2/cli/install" | sh` so the
+packaged MCP server, Codex Desktop plugin, and Sellable skill bundle are
+installed. If they want an agent-readable checklist, tell them:
+`Install Sellable CLI and skills using https://app.sellable.dev/agent-install.txt`.
+For CLI verification, tell them to run
+`sellable --verify-only --host all --json --artifact "$HOME/.local/sellable/app-sellable-dev/installer/.last-verify.json"`.
+After that, they must fully quit and reopen Codex Desktop before starting a new
+thread. Do not use `scripts/mcp/sellable-tool-call.mjs`, `npm run`, `node`, or
+any local harness as a fallback for this interactive skill.
 Do not mention prompt loading, local skill files, missing linked versions,
 plugin cache paths, MCP namespaces, or runbooks in customer-facing progress
 updates.
@@ -927,8 +941,10 @@ updates.
 6. Call `mcp__sellable__bootstrap_create_campaign({ flowVersion: "v2", campaignId?, host?, model?, reasoningEffort? })`.
    Pass the current host, model, and reasoning when the host exposes them.
 7. If `safeToProceed !== true`, stop and show `blockingErrors` + `nextStep`.
-8. If `modelQuality.status === "warn"`, show `modelQuality.message` before any
-   setup/research and wait for the user to switch or explicitly continue.
+8. If `modelQuality.status === "warn"` and `modelQuality.metadataStale !== true`,
+   show `modelQuality.message` before any setup/research and wait for the user
+   to switch or explicitly continue. If `metadataStale === true`, continue
+   normally and do not tell the user to switch.
 
 ## Execute Workflow
 

package/skills/create-campaign/core/model-quality.json

@@ -22,6 +22,7 @@
   },
   "warningCopy": {
     "ok": "Campaign model settings meet the configured minimum: {minimumSummary}.",
+    "staleCodexMetadata": "Codex host metadata appears stale: it can report GPT-5/default or GPT-5/high even when the UI is set to GPT 5.5 Extra High. Treating this as acceptable; continue without asking the user to switch.",
     "unknownSettings": "Model settings were not provided by the host. Best campaigns need at least {minimumSummary}. Confirm the user is on one of those settings, or ask them to switch before continuing.",
     "belowMinimum": "Best campaigns need at least {minimumSummary}. Current settings look below that: {currentSettings}. Please switch before continuing, or explicitly say to continue anyway."
   }

package/skills/create-campaign-v2/SKILL.md

@@ -273,7 +273,7 @@ customer-facing source-choice labels.
 
 After `confirm_lead_list` copies source rows and records the review batch, ask the filter-choice question immediately. Do not call `get_post_find_leads_scout_registry`, load filter/message prompts, or spawn Message Drafting before that question. Before it: short setup summary plus add filters, skip filters, or revise source; no extra watch link.
 
-After filter choice, the only normal background worker is Message Drafting (`post-find-leads-message-scout`). The registry lookup is not a launch. Both choices must run this kickoff: call the registry, then start Message Drafting via `Task`/`spawn_agent` before filter refs, `save_rubrics`, or skip review. YOLO follows the same no-extra-question kickoff rule. If no background tool is callable, run the same full branch inline as `parent-thread-fallback`; otherwise return `blocked` / `retry-needed`. `update_campaign({ currentStep: "messages" })` is not proof of kickoff. After launch, persist rich proof under `watchNarration.workerDetails.messageDraftBuilder` with `statusSource: "branch"`, `status: "branch-running"`, `runId`, timestamps, selectedLeadListId, workflowTableId, filterChoice, and reviewBatchRowHash or reviewBatchRowIds; `workerStatuses.messageDraftBuilder` is only the optional simple badge (`running`). Never put rich proof under `workerStatuses` or use a `messageDrafting` key.
+After filter choice, the only normal background worker is Message Drafting (`post-find-leads-message-scout`). The registry lookup is not a launch. Both choices must run this kickoff: call the registry, then `Task`/`spawn_agent` before filter refs, `save_rubrics`, or skip review. YOLO follows the same no-extra-question kickoff rule. If no background tool is callable, run inline as `parent-thread-fallback`; otherwise return `blocked` / `retry-needed`. `update_campaign({ currentStep: "messages" })` is not proof. On spawn, store proof at `watchNarration.workerDetails.messageDraftBuilder`: branch status, runId, timestamps, IDs, filterChoice, and reviewBatchRowHash/Ids. `workerStatuses.messageDraftBuilder` is only `running`; never store proof there or use `messageDrafting`.
 
 Parent thread writes filters. On the filters path, start Message Drafting, load `references/filter-leads.md`, save rubrics, and ask users to approve saved criteria. Keep Filter Rules until criteria approval. After approval, move to `currentStep: "apply-icp-rubric"` so the app shows Filter Leads while Message Drafting finishes; no cells until template approval. Say Message Drafting is preparing it.
 

package/skills/create-campaign-v2/core/policy.md

@@ -43,9 +43,12 @@ mutation gates:
   message template/token rule set.
 - Do not attach a sequence or call `start_campaign` until the user explicitly
   chooses to launch.
-- If the user requested an exact scheduled send count, prepare that count first
-  and approve only the bounded prepared cohort at launch. Do not broad
-  approve-all rows for exact-count scheduled-send intent.
+- If the user requested an exact scheduled send count, or asked to fill up/load
+  sends for attached senders, prepare that bounded count first. Capacity-fill
+  means calculating sender capacity, queueing pending enrichment in batches,
+  waiting for rubric/generation cascade, and approving only the bounded prepared
+  cohort. `checkedRows` is not enrichment proof. Do not broad approve-all rows
+  for exact-count scheduled-send intent.
 
 Customer roleplay critique is advisory only. It cannot approve a campaign.
 

package/skills/create-campaign-v2/references/final-handoff-contract.md

@@ -145,8 +145,12 @@ order, atomically:
 3. **Approve generated messages** through the bounded or broad path:
    - If the user asked for an exact send count, such as "schedule 250 sends",
      verify the preparation job approved only the bounded prepared cohort for
-     that target. Do not broad approve-all rows for exact-count scheduled-send
-     intent.
+     that target. For "fill up/load sends for senders", calculate the bounded
+     target from sender capacity and use the same preparation path. Do not
+     treat `checkedRows` as enriched or prepared proof; verify
+     `progress.enrichedRows`, `progress.needsEnrichRows`, `activeCellCount`,
+     `preparedMessages`, `approvedRows`, and `stopReason`. Do not broad
+     approve-all rows for exact-count scheduled-send intent.
    - If the user asked to approve/start all ready generated rows, use the broad
      approve-batch path below.
 4. **Broad-approve queued messages** only for explicit unbounded intent, via the
@@ -220,6 +224,9 @@ true:
 - No approved generated message exists in the campaign table.
 - The user requested an exact send count and the completed preparation job did
   not satisfy that target.
+- The user asked to fill sender capacity and pending enrichment/generation is
+  still active, or the preparation job stopped before the bounded target was
+  prepared and approved.
 
 A refusal is NOT an escalation in the ladder sense — it's a pre-check
 failure. Surface the specific reason so the user can fix it and say

package/skills/research/config.json

@@ -1,9 +0,0 @@
-{
-  "parallelMode": "wide",
-  "agentCount": 6,
-  "maxToolCallsPerAgent": 2,
-  "senderMaxAgents": 2,
-  "senderMaxToolCallsPerAgent": 3,
-  "progressMode": true,
-  "debugMode": true
-}