@hailer/mcp 1.3.21 → 1.3.32

package/.claude/skills/tool-response-verification/SKILL.md

@@ -0,0 +1,143 @@
+---
+name: tool-response-verification
+description: Verify MCP tool execution results before reporting success. Load when creating activities, updating activities, listing activities, or any MCP tool call where a silent failure could cause incorrect reporting. Covers the two distinct response envelopes (MCP wire result vs agent delegation report), create vs update response shapes, error asymmetry, empty-list shape, and MCP connection error detection.
+---
+
+# Tool Response Verification
+
+## Purpose
+
+Ensure agents verify actual tool execution results before reporting success. Prevents fabricated responses when MCP servers are disconnected or tools fail silently.
+
+---
+
+## Pattern 1: Always Check Tool Results
+
+Every MCP tool call returns a result. Read it before proceeding.
+
+## Pattern 2: Never Assume Success
+
+If you didn't see confirmation in the result, it didn't happen.
+
+## Pattern 3: Report Actual Errors
+
+If a tool returned an error, report error status — don't fabricate success.
+
+## Pattern 4: No Fabrication
+
+Never claim "created 5 activities" if you didn't see 5 activity IDs returned.
+
+## Pattern 5: Verification Checklist
+
+Before returning success:
+- Did the tool return a result (not an error)?
+- Does the result contain expected data (IDs, counts, etc.)?
+- Am I reporting data FROM the result, not data I expected?
+
+---
+
+## Pattern 6: Two Distinct Envelopes — Don't Confuse Them
+
+**MCP tool wire result** — what the tool actually returns:
+```
+{ content: [{ type: "text", text: "..." }], isError?: boolean }
+```
+Success and error both arrive as `text` content. Errors are signalled by a ❌ prefix in the text (e.g. `"❌ Activity not found"`), not by a structured status field. Activity/discussion tools do not set `isError: true` — read the text to determine outcome.
+
+**Agent delegation report** — what the agent returns to the orchestrator after verifying the tool result:
+```json
+{ "status": "success|error", "result": { ... }, "summary": "..." }
+```
+This `{status, result, summary}` shape is the agent's own output. It is NOT the MCP tool return value.
+
+---
+
+## Pattern 7: Runtime-Verified Wire Shapes for Activity Tools
+
+The following shapes are runtime-verified against testspace 2026-06-07.
+
+### Success — create_activity (single and bulk)
+
+Both single and bulk creates emit a fenced JSON block:
+```json
+{"created_ids":["<24-hex-id>", ...]}
+```
+Single create also includes a `discussionId` field alongside `created_ids`. There is NO prose-only success path for create — if you don't see `created_ids`, the create did not succeed.
+
+### Success — update_activity (single and bulk)
+
+Update success is **prose only** — there is NO `created_ids` on update responses.
+
+- Single: `✅ Successfully updated activity <id>!` + `API Response: 1`
+- Bulk: `🎉 Successfully updated N activities!` + `Total Updated: N`
+
+### Error Envelope — Asymmetry Between create and update
+
+Both create and update errors are prefixed with `❌`. There is no structured `isError`/`status` field on activity tool responses — read the text.
+
+**create_activity errors are RICH:**
+- 191 (field-level): full JSON `{"msg":"...","code":191,"details":{...},"debug":...}`
+- Zod validation: structured `**Validation Error for create_activity**` block, e.g. `name: Activity name cannot be empty` / `workflowId: Required, phaseId: Required`
+
+**update_activity errors are LOSSY:**
+- Any rejection collapses to `❌ Error updating activity: [object Object]` — the underlying message is swallowed. Verify update outcomes via read-back (`show_activity_by_id`) rather than relying on the error text.
+
+### Empty-List Shape
+
+`list_activities` with zero results returns:
+```
+✅ Found 0 activities in workflow "X"
+📊 PAGINATION INFO ... Activities on this page: 0
+[]
+```
+A human-readable string + a bare `[]` — NOT a `{activities:[]}` JSON envelope. Do not flag a bare `[]` as a parse error or an offline-server symptom.
+
+---
+
+## Pattern 8: Detect MCP Connection Errors
+
+When an MCP server is offline, tools may return unexpected text or transport-level errors.
+
+**Red flags in the text content:**
+- "0 Workflows Found" / "Activities on this page: 0" when the workspace is known to have data AND transport-layer error strings are present
+- Error strings containing: "connection", "ECONNREFUSED", "timeout", "unavailable", "ENOTFOUND"
+
+**Note:** `describe_workflows` returns a text blob like `"📋 N Workflows Found"` — there is no `workflows: []` JSON array. Read the text for the count and workflow names.
+
+**What to do:**
+
+```json
+{
+  "status": "error",
+  "result": { "tool": "tool_name", "error": "MCP server appears offline" },
+  "summary": "Cannot proceed: MCP server not responding"
+}
+```
+
+---
+
+## Examples
+
+### Correct — Verify Before Reporting
+
+```
+GOOD: Call create_activity → read result → if result.created_ids has entries → return success
+GOOD: "Created customer with ID abc123" (ID from actual result)
+```
+
+### Wrong — Fabricating Success
+
+```
+BAD: Call create_activity → immediately return {"status": "success", "created": 1}
+BAD: "I created the customer" (without checking result)
+```
+
+### Correct — Error Reporting
+
+```json
+{
+  "status": "error",
+  "result": { "tool": "create_activity", "error": "code 191: 'priority' must be a string" },
+  "summary": "Tool failed: field validation error"
+}
+```

package/CLAUDE.md

@@ -65,7 +65,7 @@ workspace/
 ├── templates/         # Document templates
 └── <WorkflowName>_<id>/
     ├── fields.ts      # Field definitions (type, label, key, options, required)
-    ├── phases.ts      # Phases (name, key, isInitial, isEndpoint, nextPhases)
+    ├── phases.ts      # Phases (name, key, isInitial, isEndpoint, possibleNextPhase)
     ├── main.ts        # Workflow config
     └── functions/     # Function field code (one file per function)
 ```
@@ -95,13 +95,27 @@ Always use `:force` — the non-force variants prompt interactively.
 |--------|---------|-------|
 | New entities | Omit `_id` — server assigns it | Don't add `_id` manually |
 | After push | `npm run pull` to get server-assigned IDs | Don't assume local IDs are final |
-| Function fields | `npm run fields-push:force` only (non-force skips `functionVariables`) | Don't use non-force |
+| Function fields | `npm run fields-push:force` — function code + `functionVariables` push from `functions/` regardless of force | non-force only prompts on deletes and hangs agents |
 | Enum imports | Verify after pull — identical hex suffixes can collide | Don't blindly trust auto-generated imports |
 | `linkedfrom` fields | Don't work in isolated-vm; use `<` backlink dependency | Don't use in function fields |
 | Function field code | Plain JavaScript only | Not TypeScript |
-| Phase transitions | Exact string match on phase name | Don't guess phase names |
+| Phase transitions | Move with the target `phaseId` (hex or key) from `describe_workflows` / enums | Don't pass the phase name string — it's not resolved |
 | Activity field values (MCP) | Date: Unix ms (`1730937600000`); Dropdown: exact string; ActivityLink: string ID; fields keyed by field ID, never label | Not ISO dates, not arrays, not field names |
 
+### New-workflow pull-wipe guard
+
+`npm run pull` after `npm run workflows-sync` **wipes** any locally-staged field/phase edits on other workflows. Safe sequence:
+
+1. `npm run workflows-sync` → check CLI output for `✨ Created: N` (N ≥ 1). If N = 0, the create was a no-op — do NOT pull.
+2. `npm run pull` — only after confirming the workflow was created.
+3. Complete the new workflow end-to-end (fields-push → phases-push) before pulling for any other reason.
+
+Never mix un-pushed field edits on existing workflows into the same batch as a new-workflow create-then-pull.
+
+### No multi-select field types
+
+Hailer fields hold **one value**. There is no multi-select for activitylink, dropdown (textpredefinedoptions), users, or teams. Writing `["id1", "id2"]` for any of these is a bug — it will silently fail or be rejected. To attach multiple linked activities, use multiple separate ActivityLink fields (or the inverse backlink). To assign multiple users, use multiple user fields.
+
 ## App Lifecycle
 
 | Step | How |
@@ -132,25 +146,37 @@ Tools not in this list are either SDK-shadowed (see matrix above) or admin/dange
 
 ## Skills
 
-Load in sub-agents via the Skill tool. Never in main context.
+Skills live in `.claude/skills/`. Load the relevant skill before starting that kind of work — it carries the patterns, gotchas, and correct formats for that domain.
+
+### Skill routing
+
+| Task / symptom | Load this skill |
+|----------------|-----------------|
+| Creating or editing workflows, fields, phases, teams, groups — or debugging push errors, field-not-visible issues, phase ordering | `sdk-ws-config-skill` |
+| Writing or debugging any function (calculated) field; code 191 compile errors; null returns; cross-workflow aggregation | `sdk-function-fields` |
+| Creating or updating activities via MCP; bulk create; field format questions; `createMany`/`updateMany`; silent write failures | `sdk-activity-patterns` |
+| Writing or debugging any insight SQL query; timestamp unit confusion (seconds vs ms); reserved column names; NULL aggregates | `sdk-insight-queries` |
+| Insight calculations beyond plain SUM/COUNT — weighted sums, CASE bucketing, conversion rates, date truncation, top-N, percentage of total | `sdk-insight-calculations` |
+| Cross-workflow insight JOINs; NULL join results; ActivityLink display values; row multiplication from one-to-many | `insight-join-patterns` |
+| Verifying MCP tool results; checking create/update actually succeeded; detecting silent failures | `tool-response-verification` |
+| Building a Hailer app UI — components, hooks, reads/writes, v1↔v2 migration, ActivityLink shapes | `hailer-app-builder` |
+| Scaffolding and publishing an app — `@hailer/create-app`, `prep-app`, publish scripts, auth, marketplace publish | `create-and-publish-app` |
+| Theme, colors, Chakra tokens, icons, layout patterns | `hailer-design-system` |
+| Adding images or media to an app | `hailer-apps-pictures` |
+| Chakra UI primitives — Select, ConfirmDialog, date pickers, state machines, REST calls from app | `hailer-app-primitives` |
+| Publish script flags, `--market`, `--create`, post-publish app registration | `publish-hailer-app` |
+| Any question the local skills don't answer — API endpoint arguments/returns, RPC names, REST routes, SDK/CLI usage, end-user how-tos: search the official docs at docs.hailer.com | `hailer-docs-search` |
+
+### Field-by-key resolver (apps)
 
-| Skill | When |
-|-------|------|
-| `sdk-ws-config-skill` | Workflows, fields, phases, teams, groups |
-| `sdk-function-fields` | Calculated fields, nameFunction |
-| `sdk-activity-patterns` | Creating/updating activities via MCP |
-| `sdk-insight-queries` | SQL insight queries |
-| `hailer-app-builder` | Building Hailer apps (React/Chakra) |
-| `hailer-design-system` | Theme, colors, icons, layout |
-| `create-and-publish-app` | Scaffolding + publishing apps (dev/prod/marketplace) |
-| `testing-patterns` | Vitest/playwright tests |
+In app code, reference fields by their **key** via the generated field resolver — never hardcode hex field IDs. A field with `key: "status"` is accessible as `activity.fields.status` (with `returnFlat: true`) or via the resolver. This keeps apps workspace-portable if fields are recreated.
 
-## Commands
+Fields without a `key` fall back to hex ID lookup. Assign meaningful keys when creating fields.
 
-`/command <param>` — angle brackets = required
+## Design before building
 
-**Essential:** `/save`, `/handoff`, `/prd`, `/autoplan`, `/ws-pull`
+Before building a workspace, load `hailer-workflow-archetypes` to pick the structural shape (Register, Pipeline, Header+LineItems, etc.), then `hailer-design-patterns` for parent/child lifecycle, row progression, and integration patterns. Design the structure first — flat field-piles without archetype thinking cause expensive rework later.
 
-**Squads** (drive the native Workflow tool): `/review-squad [target]`, `/debug-squad "<bug>"`
+## Build strategy
 
-Build apps with a single agent + the `hailer-app-builder` / `create-and-publish-app` skills — faster than fanning out a squad (no shared-contract prep, no cross-agent integration fixups).
+Build apps with a single agent (`agent-giuseppe-app-builder`) + the `hailer-app-builder` / `create-and-publish-app` skills — a focused agent dispatch is more efficient than multi-agent fan-out for app builds (no shared-contract prep, no cross-agent integration fixups).

package/dist/bot/bot.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"bot.d.ts","sourceRoot":"","sources":["../../src/bot/bot.ts"],"names":[],"mappings":"AAAA;;;;;;;;;GASG;AAIH,OAAO,EAAE,YAAY,EAAa,MAAM,sBAAsB,CAAC;AAwB/D,OAAO,EAAkC,YAAY,EAAE,MAAM,kBAAkB,CAAC;AAChF,OAAO,EAEH,SAAS,EAMZ,MAAM,eAAe,CAAC;AAIvB,iEAAiE;AACjE,UAAU,aAAa;IACrB,yBAAyB,CAAC,WAAW,EAAE,MAAM,GAAG,GAAG,CAAC,MAAM,CAAC,CAAC;CAC7D;AAED,mEAAmE;AACnE,MAAM,WAAW,UAAU;IACzB,eAAe,EAAE,MAAM,CAAC;IACxB,YAAY,EAAE,YAAY,CAAC;IAC3B,UAAU,CAAC,EAAE,aAAa,CAAC;CAC5B;AA2HD;;;;;GAKG;AACH,qBAAa,iBAAkB,SAAQ,KAAK;IACrB,UAAU,EAAE,MAAM;IAAS,WAAW,EAAE,MAAM;gBAA9C,UAAU,EAAE,MAAM,EAAS,WAAW,EAAE,MAAM;CAIpE;AA4KD,qBAAa,GAAG;IACZ,OAAO,CAAC,MAAM,CAAS;IACvB,OAAO,CAAC,aAAa,CAAsB;IAC3C,OAAO,CAAC,MAAM,CAA6B;IAC3C,OAAO,CAAC,SAAS,CAA0B;IAC3C,OAAO,CAAC,YAAY,CAAe;IACnC,OAAO,CAAC,WAAW,CAA4B;IAG/C,OAAO,CAAC,IAAI,CAAyC;IACrD,OAAO,CAAC,cAAc,CAA+B;IACrD,OAAO,CAAC,iBAAiB,CAAM;IAC/B,OAAO,CAAC,MAAM,CAAM;IACpB,OAAO,CAAC,YAAY,CAAqB;IAGzC,OAAO,CAAC,aAAa,CAAqB;IAC1C,OAAO,CAAC,aAAa,CAA0B;IAG/C,OAAO,CAAC,YAAY,CAAe;IACnC,OAAO,CAAC,WAAW,CAAiB;IACpC,OAAO,CAAC,gBAAgB,CAAmB;IAC3C,OAAO,CAAC,mBAAmB,CAAoC;IAC/D,OAAO,CAAC,iBAAiB,CAAkC;IAC3D,OAAO,CAAC,gBAAgB,CAAwC;IAChE,OAAO,CAAC,eAAe,CAAuC;IAC9D,OAAO,CAAC,YAAY,CAAoC;IACxD,OAAO,CAAC,SAAS,CAAgC;IACjD,OAAO,CAAC,QAAQ,CAAyB;IACzC,OAAO,CAAC,WAAW,CAAiC;IAEpD;;;;OAIG;IACH,OAAO,CAAC,cAAc,CAA6B;IACnD,OAAO,CAAC,MAAM,CAAC,QAAQ,CAAC,WAAW,CAIhC;IACH,OAAO,CAAC,MAAM,CAAC,QAAQ,CAAC,qBAAqB,CAM1C;IAGH,OAAO,CAAC,aAAa,CAA0D;IAC/E,OAAO,CAAC,qBAAqB,CAA6B;IAC1D,OAAO,CAAC,uBAAuB,CAA6B;IAC5D,OAAO,CAAC,sBAAsB,CAA6B;IAC3D,OAAO,CAAC,UAAU,CAAS;IAE3B,OAAO,CAAC,MAAM,CAAoB;IAClC,OAAO,CAAC,eAAe,CAAS;IAChC,OAAO,CAAC,UAAU,CAAuB;gBAE7B,MAAM,EAAE,SAAS,EAAE,OAAO,EAAE,UAAU;IAqBlD,IAAI,KAAK,IAAI,MAAM,CAElB;IACD,IAAI,QAAQ,IAAI,MAAM,CAErB;IACD,IAAI,WAAW,IAAI,MAAM,CAExB;IACD,IAAI,SAAS,IAAI,OAAO,CAEvB;IACD,IAAI,WAAW,IAAI,MAAM,GAAG,SAAS,CAEpC;IACD,IAAI,SAAS,IAAI,MAAM,CAEtB;IAED,kBAAkB,CAAC,MAAM,EAAE,MAAM,GAAG,SAAS,GAAG,IAAI;IAKpD,kBAAkB,CAAC,IAAI,EAAE,YAAY,GAAG,SAAS,GAAG,IAAI;IAKxD,gBAAgB,CAAC,UAAU,EAAE,OAAO,GAAG,SAAS,GAAG,IAAI;IAOjD,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;IAoM5B;;;;;;;;;;KAUC;YACa,4BAA4B;IAoB1C;;;;;;;;;;KAUC;YACa,qBAAqB;IAuNnC;;;;;;KAMC;IACK,OAAO,IAAI,OAAO,CAAC,IAAI,CAAC;IAI9B;;;;KAIC;IACD,OAAO,CAAC,mBAAmB;IA0BrB,IAAI,IAAI,OAAO,CAAC,IAAI,CAAC;YA0Bb,YAAY;YAiIZ,iBAAiB;YAoGjB,cAAc;IA4C5B,8FAA8F;IAC9F,OAAO,CAAC,MAAM,CAAC,QAAQ,CAAC,cAAc,CAA8H;YAEtJ,YAAY;IAqD1B,OAAO,CAAC,gBAAgB;YAsBV,UAAU;YAuJV,YAAY;IAqL1B,OAAO,CAAC,qBAAqB;YAmCf,sBAAsB;YAmJtB,qBAAqB;YAqBrB,cAAc;YA4Cd,oBAAoB;IAiBlC,OAAO,CAAC,yBAAyB;IAcjC,OAAO,CAAC,aAAa;IAOrB,OAAO,CAAC,iBAAiB;YAiBX,WAAW;IAazB,OAAO,CAAC,cAAc;IAOtB,oGAAoG;YACtF,uBAAuB;IAerC,OAAO,CAAC,eAAe;CAiB1B"}
+{"version":3,"file":"bot.d.ts","sourceRoot":"","sources":["../../src/bot/bot.ts"],"names":[],"mappings":"AAAA;;;;;;;;;GASG;AAIH,OAAO,EAAE,YAAY,EAAa,MAAM,sBAAsB,CAAC;AAwB/D,OAAO,EAAkC,YAAY,EAAE,MAAM,kBAAkB,CAAC;AAChF,OAAO,EAEH,SAAS,EAMZ,MAAM,eAAe,CAAC;AAIvB,iEAAiE;AACjE,UAAU,aAAa;IACrB,yBAAyB,CAAC,WAAW,EAAE,MAAM,GAAG,GAAG,CAAC,MAAM,CAAC,CAAC;CAC7D;AAED,mEAAmE;AACnE,MAAM,WAAW,UAAU;IACzB,eAAe,EAAE,MAAM,CAAC;IACxB,YAAY,EAAE,YAAY,CAAC;IAC3B,UAAU,CAAC,EAAE,aAAa,CAAC;CAC5B;AA0ID;;;;;GAKG;AACH,qBAAa,iBAAkB,SAAQ,KAAK;IACrB,UAAU,EAAE,MAAM;IAAS,WAAW,EAAE,MAAM;gBAA9C,UAAU,EAAE,MAAM,EAAS,WAAW,EAAE,MAAM;CAIpE;AA4KD,qBAAa,GAAG;IACZ,OAAO,CAAC,MAAM,CAAS;IACvB,OAAO,CAAC,aAAa,CAAsB;IAC3C,OAAO,CAAC,MAAM,CAA6B;IAC3C,OAAO,CAAC,SAAS,CAA0B;IAC3C,OAAO,CAAC,YAAY,CAAe;IACnC,OAAO,CAAC,WAAW,CAA4B;IAG/C,OAAO,CAAC,IAAI,CAAyC;IACrD,OAAO,CAAC,cAAc,CAA+B;IACrD,OAAO,CAAC,iBAAiB,CAAM;IAC/B,OAAO,CAAC,MAAM,CAAM;IACpB,OAAO,CAAC,YAAY,CAAqB;IAGzC,OAAO,CAAC,aAAa,CAAqB;IAC1C,OAAO,CAAC,aAAa,CAA0B;IAG/C,OAAO,CAAC,YAAY,CAAe;IACnC,OAAO,CAAC,WAAW,CAAiB;IACpC,OAAO,CAAC,gBAAgB,CAAmB;IAC3C,OAAO,CAAC,mBAAmB,CAAoC;IAC/D,OAAO,CAAC,iBAAiB,CAAkC;IAC3D,OAAO,CAAC,gBAAgB,CAAwC;IAChE,OAAO,CAAC,eAAe,CAAuC;IAC9D,OAAO,CAAC,YAAY,CAAoC;IACxD,OAAO,CAAC,SAAS,CAAgC;IACjD,OAAO,CAAC,QAAQ,CAAyB;IACzC,OAAO,CAAC,WAAW,CAAiC;IAEpD;;;;OAIG;IACH,OAAO,CAAC,cAAc,CAA6B;IACnD,OAAO,CAAC,MAAM,CAAC,QAAQ,CAAC,WAAW,CAIhC;IACH,OAAO,CAAC,MAAM,CAAC,QAAQ,CAAC,qBAAqB,CAM1C;IAGH,OAAO,CAAC,aAAa,CAA0D;IAC/E,OAAO,CAAC,qBAAqB,CAA6B;IAC1D,OAAO,CAAC,uBAAuB,CAA6B;IAC5D,OAAO,CAAC,sBAAsB,CAA6B;IAC3D,OAAO,CAAC,UAAU,CAAS;IAE3B,OAAO,CAAC,MAAM,CAAoB;IAClC,OAAO,CAAC,eAAe,CAAS;IAChC,OAAO,CAAC,UAAU,CAAuB;gBAE7B,MAAM,EAAE,SAAS,EAAE,OAAO,EAAE,UAAU;IAqBlD,IAAI,KAAK,IAAI,MAAM,CAElB;IACD,IAAI,QAAQ,IAAI,MAAM,CAErB;IACD,IAAI,WAAW,IAAI,MAAM,CAExB;IACD,IAAI,SAAS,IAAI,OAAO,CAEvB;IACD,IAAI,WAAW,IAAI,MAAM,GAAG,SAAS,CAEpC;IACD,IAAI,SAAS,IAAI,MAAM,CAEtB;IAED,kBAAkB,CAAC,MAAM,EAAE,MAAM,GAAG,SAAS,GAAG,IAAI;IAKpD,kBAAkB,CAAC,IAAI,EAAE,YAAY,GAAG,SAAS,GAAG,IAAI;IAKxD,gBAAgB,CAAC,UAAU,EAAE,OAAO,GAAG,SAAS,GAAG,IAAI;IAOjD,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;IAoM5B;;;;;;;;;;KAUC;YACa,4BAA4B;IAoB1C;;;;;;;;;;KAUC;YACa,qBAAqB;IAuNnC;;;;;;KAMC;IACK,OAAO,IAAI,OAAO,CAAC,IAAI,CAAC;IAI9B;;;;KAIC;IACD,OAAO,CAAC,mBAAmB;IA0BrB,IAAI,IAAI,OAAO,CAAC,IAAI,CAAC;YA0Bb,YAAY;YAiIZ,iBAAiB;YAoGjB,cAAc;IA4C5B,8FAA8F;IAC9F,OAAO,CAAC,MAAM,CAAC,QAAQ,CAAC,cAAc,CAA8H;YAEtJ,YAAY;IAqD1B,OAAO,CAAC,gBAAgB;YAsBV,UAAU;YAuJV,YAAY;IAqL1B,OAAO,CAAC,qBAAqB;YAmCf,sBAAsB;YAmJtB,qBAAqB;YAqBrB,cAAc;YA4Cd,oBAAoB;IAiBlC,OAAO,CAAC,yBAAyB;IAcjC,OAAO,CAAC,aAAa;IAOrB,OAAO,CAAC,iBAAiB;YAiBX,WAAW;IAazB,OAAO,CAAC,cAAc;IAOtB,oGAAoG;YACtF,uBAAuB;IAerC,OAAO,CAAC,eAAe;CAiB1B"}

package/dist/bot/bot.js

@@ -119,8 +119,28 @@ function parseSeedTurns(raw, log) {
             const namesHint = templateNames.length > 0
                 ? `\n\n[INSTALL THESE PRODUCTS BY NAME: ${templateNames.join(', ')}]`
                 : '';
+            // Append discovery brief if available
+            const briefRaw = parsed.discoveryBrief;
+            let briefHint = '';
+            if (briefRaw && typeof briefRaw === 'object') {
+                const b = briefRaw;
+                const parts = [];
+                if (b.industry)
+                    parts.push(`Industry: ${b.industry}`);
+                if (b.job)
+                    parts.push(`Needs: ${b.job}`);
+                if (b.today)
+                    parts.push(`Current tools: ${b.today}`);
+                if (b.door)
+                    parts.push(`Experience: ${b.door}`);
+                if (b.must_haves)
+                    parts.push(`Must-haves: ${b.must_haves}`);
+                if (parts.length > 0) {
+                    briefHint = `\n\n[DISCOVERY BRIEF]\n${parts.join('\n')}`;
+                }
+            }
             return {
-                turns: [{ role: 'user', content: `[Demo conversation summary]\n${summary}${namesHint}` }],
+                turns: [{ role: 'user', content: `[Demo conversation summary]\n${summary}${namesHint}${briefHint}` }],
                 viewedSlides,
                 templateIds,
                 templateNames,