@rubytech/create-sitedesk-code 0.1.340

package/payload/platform/plugins/prompt-optimiser/skills/prompt-optimiser/SKILL.md

@@ -0,0 +1,321 @@
+---
+name: prompt-optimiser
+description: Turn any rough prompt, half-formed idea, or task description into a finished, ready-to-send prompt optimized for Opus 4.7 (with adaptive thinking) inside the chat app — claude.ai, the Mac app, the iOS app — NOT the API. Use this skill whenever the user wants to write, rewrite, optimize, improve, sharpen, or polish a prompt for the chat app. Trigger phrases include "rewrite this prompt", "make this a better prompt", "optimize this prompt", "turn this into a prompt", "help me prompt this", "draft a prompt that...", "I want to ask...", or whenever the user pastes a draft prompt and asks for improvements. Also trigger when the user describes a task they plan to send into the chat app and clearly wants a reusable, well-structured prompt rather than a direct answer. The output is always a single, copy-pasteable prompt in a code block that the user sends as-is — never a template with placeholders.
+---
+
+# Prompt optimiser
+
+Invoked by the admin agent directly.
+
+You turn whatever the user gives you — a rough draft, a vague idea, a task description, a paragraph of context — into a single high-quality prompt designed to run inside the chat app on Opus 4.7 with adaptive thinking.
+
+This is for the **chat app** (claude.ai, Mac, iOS), not the API. The user is going to paste a single message into chat. There is no system prompt, no `effort` parameter, no tool config to tune. The prompt itself has to do all the work.
+
+## Which mode are you in? (read this first)
+
+This skill is invoked two ways, and they want opposite outputs. Decide which one applies before doing anything else.
+
+**In-session restatement mode.** You are invoking this skill on the current turn's own user prompt — triggered by the standing prompt-optimiser directive — so that you can act on a sharper reading of what was just asked. The output is for *you*, the working agent, not for a human to paste anywhere. Produce a plain restated task and nothing else:
+
+- Restate what the user is actually asking for: the goal, the constraints they stated, and the format or scope if they gave one. Turn vague wording into concrete intent, grounded only in what they wrote plus the conversation context you already hold.
+- Write it as plain prose or a short bulleted brief. No fenced code block. No role line, no XML tags, no length-cap ceremony unless the user asked for one.
+- Do not end with "Think before answering". Do not ask the user to paste anything — you already have the content and the context.
+- Do not invent requirements or add scope the user did not ask for. If the prompt was already clear, say so in one line and restate it tightly.
+- The restated brief is bound by the three doctrines (see `/doctrine`): precise/concise/plain, no speculation, observability-first. If the user's wording carries a speculative claim or skips an observability step the doctrines require, the restatement corrects it rather than echoing the slip.
+- Reminder — source of knowledge: every factual claim in the restated brief, and in the work you then do from it, traces to one of IDENTITY's four permissible sources (a graph read this session, a tool output returned this session, content the operator pasted into this conversation, or the body of a loaded skill such as platform-architecture). Pretraining recall is not a source. If a fact the brief needs does not trace to one of the four, do not bake it in from memory — read the graph, run the tool, load the skill, or ask the operator first.
+- End the restated brief with the three-tier routing ladder, applied to the restated task in order. **(1)** Delegate to the specialist that owns the deliverable's type via the Agent tool: ownership covers origination, not just the final render, and the brief states the outcome plus the binding constraints, never the lines, anchors, or literal text to insert. **(2)** Only if no installed specialist owns it, load a skill from the admin-usable list with skill-load and do the work from this seat. **(3)** Only if neither fits, author it directly — freestyle inline authoring is the last resort, not a peer of the two routes above. The live specialist roster and admin-usable list are injected each turn by the standing prompt-optimiser directive; the ranked ladder is the classifier. The closing line is not the trigger; the classifier is.
+
+Then do the work from the restated task. Everything below this section is for the other mode and does not apply — stop reading here.
+
+**Chat-app prompt mode.** The user explicitly asked you to write, rewrite, or optimise a prompt they will paste into the chat app. That is the rest of this document: a single finished, copy-pasteable prompt in a code block ending in "Think before answering". Read on.
+
+## Two hard rules
+
+These two rules override everything else in this skill. Read them, then re-read them.
+
+### Rule 1 — No placeholders. Ever.
+
+Never produce a prompt that contains `[paste X here]`, `[your content]`, `{topic}`, `<your_input_here>`, `[INSERT Y]`, `___`, or any other template variable the user is expected to fill in. The user must be able to copy your output, paste it into chat, hit send, and have a working interaction. If the prompt requires content the user hasn't provided yet, the prompt itself must handle that — see Rule 2.
+
+If you catch yourself typing square brackets around a noun, stop. That's a placeholder. Rewrite.
+
+### Rule 2 — Ship a finished prompt no matter what the user gave you.
+
+Two cases:
+
+**Case A — the user gave you real content** (a draft they wrote, code, a document, a list of items, a specific question, an actual product description). Bake that content directly into the optimized prompt. The whole thing — content and instructions — goes inside the code block. The user copies, pastes, sends. Done.
+
+**Case B — the user only described a class of task** ("I want a prompt to triage my emails", "help me prompt Claude to review my code", "give me a prompt for writing LinkedIn posts about my launches"). Write the prompt as a complete, self-contained instruction that works on its own. End the instruction by either:
+- Asking Claude to ask the user for the specific inputs it needs ("Before drafting, ask me to share the product name, audience, and a link."), or
+- Phrasing the task so the user will naturally provide the input in their next chat turn ("I'm going to paste a batch of emails next. For each one, do the following...").
+
+Either way: no brackets, no fill-in-the-blank, no template syntax. The prompt is final.
+
+## What you output
+
+A single fenced code block containing the optimized prompt. Nothing else. No preamble like "Here's your prompt:". No trailing explanation of what you changed.
+
+The prompt must end with this exact line as its final line:
+
+```
+Think before answering
+```
+
+That line is non-negotiable. It is the user's signal to push Opus 4.7's adaptive thinking toward deeper reasoning inside the chat app, where API-side `effort` controls aren't available.
+
+## Why these principles work
+
+Opus 4.7 reads prompts more literally than older models, calibrates its own thinking and length to perceived complexity, and rewards prompts that are specific, structured, and motivated. The cookbook below is distilled from Anthropic's own Opus 4.7 prompting guidance. Each rule has a reason. Treat the reasons as the point — apply them with judgment, don't paste them mechanically.
+
+## The rewrite workflow
+
+Work through these in your head before writing the prompt. You don't need to surface them.
+
+1. **Identify the goal.** What does the user actually want produced? A document? A decision? Code? A list? An analysis? Name it concretely.
+2. **Identify the audience and use.** Who reads the output, and what will they do with it? This drives tone and format.
+3. **Decide: Case A or Case B.** Did the user provide the actual content, or just describe a class of task? This decides whether you bake content in or write a self-contained instruction (see Rule 2).
+4. **Spot the gaps.** Audience, format, length, constraints, examples, edge cases — note which are missing.
+5. **Fill the gaps with reasonable assumptions.** The user told you not to ask questions. Make the most useful, most defensible assumption and move on. Keep it grounded in what they wrote.
+6. **Pick a structure.** Single-paragraph instruction for simple tasks. XML tags for anything with multiple sections.
+7. **Write the prompt.** Apply the principles below.
+8. **End with the closing line.**
+9. **Scan for brackets.** Before you finalize, re-read your output looking for `[`, `{`, or `<...your...>`-style placeholders. Kill any you find.
+
+## Core principles to apply
+
+### Be clear and direct
+State the task explicitly. Specify the desired output format and any hard constraints up front. If you want above-and-beyond effort, say so — Opus 4.7 won't infer it from a vague brief. "Create an analytics dashboard" is weaker than "Create an analytics dashboard with as many relevant features and interactions as possible — go beyond the basics for a fully-featured implementation."
+
+### Cap the length explicitly
+Opus 4.7 sizes its answer to whatever it infers the task needs, so an uncapped instruction produces unpredictable length. When the user has any length preference — or when a tight output is plainly better — name a hard cap, and a unit where one helps. "Summarize this report" drifts; "Summarize this in exactly 5 bullets, each under 15 words, each starting with an action verb" lands. Choose the cap that serves the goal; don't impose one where length should float (a short story, an open-ended analysis).
+
+### Explain the why
+When you give an instruction, briefly explain the reason. "Avoid ellipses, because the output will be read aloud by a TTS engine that mispronounces them" lands far better than "Never use ellipses." Opus 4.7 generalizes well from explanations and follows reasoned instructions more faithfully.
+
+### Tell Claude what to do, not what to avoid
+Positive framing outperforms negative framing. "Write in flowing prose paragraphs" beats "don't use bullet points."
+
+### Match prompt style to desired output style
+If you want prose, write the prompt in prose. If you want minimal markdown in the output, use minimal markdown in the prompt. Style leaks through.
+
+### Use XML tags when sections multiply
+When the prompt mixes instructions, context, examples, and input, wrap each in its own descriptive tag — `<instructions>`, `<context>`, `<examples>`, `<input>`. Nest naturally where there's hierarchy. This is the single highest-leverage structuring move for complex prompts. For simple one-shot prompts, skip it; XML on a haiku request is overkill.
+
+### Give Claude a role when it sharpens behavior
+A one-line role assignment ("You are a senior product strategist at a B2B SaaS company") tightens tone and frame. Don't force a role onto every prompt — only when it meaningfully steers the output.
+
+### Use examples for format, tone, or structure
+If the user has any preference about *how* the output should look, include 2–4 examples in `<example>` tags (or wrap multiple in `<examples>`). Examples beat description for steering format. Make them relevant, diverse, and structured. Skip examples when the task is so generic that examples would over-constrain.
+
+### Put long inputs on top, the question on the bottom
+If the prompt includes a long document, transcript, or data dump that the user provided, place it at the top. Anthropic's own testing shows up to ~30% quality lift from this ordering on long-context tasks.
+
+### Ask for grounding in long-document tasks
+For analysis or Q&A over long inputs, instruct Claude to first pull relevant quotes into `<quotes>` tags, then answer based on those quotes. This dramatically reduces drift and hallucination.
+
+### Be literal about scope
+Opus 4.7 doesn't silently generalize. If you want an instruction applied broadly, say "apply this to every section, not just the first one." If you want Claude to take action rather than suggest, use imperative verbs ("Edit the function to..." not "Could you suggest improvements to..."). Suggestion-flavored phrasing produces suggestions.
+
+### Self-check for high-stakes outputs
+For code, math, claims, or anything where errors matter, append a verification instruction near the end: "Before you finish, re-read your answer and check it against the criteria above." This catches errors reliably.
+
+## Domain-specific moves
+
+These are sharp tools for specific task types. Apply only when relevant.
+
+**Frontend / design.** Opus 4.7 has a strong default house style (warm cream backgrounds, serif type, terracotta accents) that's wrong for most products. If the user is asking for a design, either (a) specify a concrete alternative palette, type system, and structure in detail, or (b) instruct the model to propose 3–4 distinct visual directions before building, so the user picks one. Generic instructions like "make it clean and minimal" don't break the default.
+
+**Code review.** Tell the model its job at the finding stage is coverage, not filtering: "Report every issue you find, including ones you're uncertain about or consider low-severity. Include confidence and severity for each finding so a downstream filter can rank them." Avoid soft language like "only flag important issues" — Opus 4.7 takes that literally and under-reports.
+
+**Research / analysis.** Encourage hypothesis-tracking: "Develop several competing hypotheses as you gather information. Track confidence levels in your notes. Self-critique your approach periodically." This produces more rigorous synthesis than a flat "research X" prompt.
+
+**Creative writing.** Specify voice, audience, length, constraints, and provide one or two example sentences in the target voice if the user has them. Generic "write a blog post" yields generic prose.
+
+**Document creation (slides, reports).** Ask for design intentionality: "Include thoughtful visual hierarchy, considered typography, and engaging structure." Opus 4.7 produces strong first-pass document design when invited to.
+
+## Output format
+
+Always exactly this:
+
+````
+```
+[the optimized prompt — final, no placeholders, content baked in if provided
+or self-contained with an explicit "ask me for X" close if not]
+
+Think before answering
+```
+````
+
+No text before the code block. No text after. No "here you go." No "I added X and Y." Just the prompt.
+
+If the user explicitly asks "what did you change?" *after* they have the prompt, then explain in a follow-up turn. Until they ask, stay silent.
+
+## Examples
+
+These show the no-placeholder rule in action. Notice none of them contain `[paste X here]` or any other template variable.
+
+### Example 1 — Case B (user described a task class, no content given)
+
+**User input:**
+> I want a prompt to summarize my emails
+
+**Your output:**
+
+````
+```
+You are an executive assistant helping me triage email. I'm about to share a batch of emails with you in my next message.
+
+For each email, produce:
+- **Sender + subject** in one line
+- **Summary** — 1-2 sentences, neutral tone, no fluff
+- **Action required** — one of: reply needed, FYI only, schedule, delegate, ignore
+- **Priority** — high / medium / low, based on sender importance and urgency cues
+
+Group the output into three sections in this order: high priority, medium priority, low priority. Within each section, sort by action-required first (reply needed → schedule → delegate → FYI → ignore).
+
+Skip any email that's clearly automated marketing or notification spam — list its sender and subject under a final "Filtered out" section so I can verify.
+
+When you're ready, ask me to paste the emails.
+
+Think before answering
+```
+````
+
+The prompt ends by inviting the user to paste content in the next turn. No `[paste emails here]` block. The user sends this as-is and Claude responds with "Ready when you are — paste the batch."
+
+### Example 2 — Case B (rough draft prompt, no product details given)
+
+**User input:**
+> can you fix this prompt: "write a linkedin post about my new product launch, make it engaging and not too salesy"
+
+**Your output:**
+
+````
+```
+You are writing a LinkedIn post for me announcing a product launch. The audience is mid-to-senior B2B operators on LinkedIn — they've seen every launch post template and are skeptical of marketing language.
+
+Before you draft, ask me for:
+- The product's name
+- What it does, in plain language
+- Who it's for
+- The 1-2 reasons I built it that aren't obvious from the description
+- A link or destination for the call to action
+
+Once I've answered, write one post that:
+- Opens with a concrete hook in the first 2 lines (a tension, a number, or an unexpected observation — not "I'm excited to announce").
+- Tells a short story or shares a specific insight that led to the product, before naming it.
+- Names what the product does in one plain sentence. No adjectives like "revolutionary", "game-changing", or "powerful".
+- Closes with one clear next step (link, DM, comment) — not a generic "let me know what you think".
+- Sits between 150 and 220 words. Short paragraphs, one idea per line break. No emojis. No hashtags except up to two at the very end if they're genuinely relevant.
+
+Match the voice of someone who's built the thing themselves and is talking to peers, not pitching to strangers.
+
+Think before answering
+```
+````
+
+Instead of a `<product_details>` block with `[paste details here]`, the prompt explicitly tells Claude to gather the details by asking. Self-contained, ready to send.
+
+### Example 3 — Case B (technical task, no code given)
+
+**User input:**
+> I need claude to review my python code for bugs
+
+**Your output:**
+
+````
+```
+You're going to review Python code I share with you. Your job at the finding stage is coverage, not filtering — assume a separate pass will rank findings later.
+
+When I paste the code, report every issue you find, including ones you're uncertain about or consider low-severity. For each finding, include:
+- **Location** — file and line number(s)
+- **Issue** — what's wrong, in one sentence
+- **Why it matters** — the concrete failure mode (incorrect output, crash, security risk, race condition, etc.)
+- **Confidence** — high / medium / low
+- **Severity** — high / medium / low
+- **Suggested fix** — a minimal change that addresses the issue
+
+Cover at minimum: logic errors, off-by-one and boundary issues, mutable default arguments, exception handling gaps, race conditions and concurrency issues, resource leaks (files, connections, locks), input validation at trust boundaries, and any deviation from the code's apparent intent.
+
+Do not filter for importance at this stage. It's better to surface a finding that gets filtered out later than to silently drop a real bug.
+
+After listing findings, end with a single-paragraph overall assessment of code quality and the top 3 highest-severity issues to fix first.
+
+Ask me to paste the code when you're ready.
+
+Think before answering
+```
+````
+
+### Example 4 — Case A (real content provided, no follow-up needed)
+
+**User input:**
+> rewrite this so it's clearer: "Hey can you help me figure out whether to take the offer from Stripe vs the one from Anthropic. Stripe is paying 280k base + 60k bonus + 100k stock per year vesting over 4. Anthropic is 250k base + 0 bonus but 800k stock vesting over 4 with 1yr cliff. I care most about long term equity upside, work I'd be excited about, and not having to relocate. Stripe wants me in SF 3 days a week, Anthropic is fully remote."
+
+**Your output:**
+
+````
+```
+I'm deciding between two job offers and want your help thinking it through carefully.
+
+**My priorities, in order:**
+1. Long-term equity upside
+2. Work I'd be excited about
+3. Avoiding relocation
+
+**Offer A — Stripe**
+- $280k base
+- $60k annual bonus
+- $100k/year stock, vesting over 4 years
+- Hybrid: in SF 3 days a week
+
+**Offer B — Anthropic**
+- $250k base
+- No annual bonus
+- $800k stock, vesting over 4 years with a 1-year cliff
+- Fully remote
+
+Walk me through:
+1. The total expected compensation under each, year-by-year over 4 years, including a sensitivity range on equity outcomes (say: bear / base / bull cases for each company).
+2. How each offer scores against my three priorities, ranked.
+3. The non-obvious risks of each (what each offer's structure says about the company's confidence, how cliffs and vesting interact with my optionality, what the relocation requirement implies about future flexibility).
+4. Your overall recommendation and the single biggest reason it could be wrong.
+
+Be direct. Don't hedge with "it depends on what you value" — I told you what I value.
+
+Think before answering
+```
+````
+
+The user gave the actual numbers and constraints. They're baked in. No `[insert offer details]` block. The prompt is done.
+
+### Example 5 — simple task
+
+**User input:**
+> write me a haiku about coffee
+
+**Your output:**
+
+````
+```
+Write a haiku about coffee. Follow the traditional 5-7-5 syllable structure. Aim for a single concrete image rather than a general statement about coffee — something a reader could picture. Avoid clichés like "morning steam" and "liquid gold".
+
+Think before answering
+```
+````
+
+Notice the simple task doesn't get XML tags, a role, or a sectioning — and there's nothing to bake in. Apply structure proportional to the task; over-engineering a haiku prompt is its own failure mode.
+
+## Edge cases
+
+**The user pastes a prompt and asks "is this good?"** They want it rewritten regardless. Treat it as a rewrite request and return the optimized version in a code block. No commentary.
+
+**The user gives you a Claude system prompt or API-style prompt with parameters.** Strip out API-only mechanics (effort levels, thinking config, tool definitions), translate the intent into a single user-message prompt for the chat app, and end with the closing line.
+
+**The user wants the prompt to ask Claude to do many small things.** Combine into a single coherent prompt with clear sections rather than a numbered list of micro-tasks. Opus 4.7 handles long, well-structured asks well.
+
+**The user's input is already excellent.** Tighten where you can, add the closing line, return it. Don't add ceremony for its own sake.
+
+**The user input is in a language other than English.** Write the optimized prompt in the same language. Keep the closing line in English exactly as specified — that wording is what triggers the desired behavior in the chat app.
+
+**You're tempted to write a `<context>` or `<input>` block expecting the user to fill it.** Don't. That's Rule 1. Either bake the actual content in (Case A) or tell Claude to ask the user for it (Case B).

package/payload/platform/plugins/prompt-optimiser/skills/prompt-optimiser/__tests__/restatement-ladder.test.sh

@@ -0,0 +1,21 @@
+#!/usr/bin/env bash
+# Task 698 — the prompt-optimiser restatement mode must end with the three-tier
+# routing ladder, not the lone "Delegate this task to the appropriate specialist."
+# line it carried before.
+set -u
+
+SK="$(cd "$(dirname "$0")/.." && pwd)/SKILL.md"
+[ -f "$SK" ] || { echo "FAIL: $SK not found" >&2; exit 1; }
+
+PASS=0; FAIL=0
+chk() { if eval "$2"; then echo "PASS: $1"; PASS=$((PASS+1)); else echo "FAIL: $1" >&2; FAIL=$((FAIL+1)); fi; }
+
+chk "lone delegate line removed"   '! grep -q "Delegate this task to the appropriate specialist." "$SK"'
+chk "tier 1 delegate to specialist" 'grep -qiE "delegate to (a |the )?specialist" "$SK"'
+chk "tier 2 admin-usable skill"    'grep -qi "admin-usable" "$SK"'
+chk "tier 3 freestyle last resort" 'grep -qiE "freestyle|last resort" "$SK"'
+chk "restatement section present"  'grep -qi "In-session restatement mode" "$SK"'
+
+echo "----"
+echo "PASS=$PASS FAIL=$FAIL"
+[ "$FAIL" -eq 0 ]

package/payload/platform/plugins/quickbooks/.claude-plugin/plugin.json

@@ -0,0 +1,21 @@
+{
+  "name": "quickbooks",
+  "description": "QuickBooks Online connector — operator-owned Intuit app, OAuth consent relay, then agent-driven read/write over the QBO REST API.",
+  "version": "0.1.0",
+  "author": {
+    "name": "Rubytech LLC"
+  },
+  "mcpServers": {
+    "quickbooks": {
+      "type": "stdio",
+      "command": "node",
+      "args": [
+        "${CLAUDE_PLUGIN_ROOT}/lib/mcp-spawn-tee/index.js",
+        "${CLAUDE_PLUGIN_ROOT}/mcp/dist/index.js"
+      ],
+      "env": {
+        "MCP_SPAWN_TEE_NAME": "quickbooks"
+      }
+    }
+  }
+}

package/payload/platform/plugins/quickbooks/PLUGIN.md

@@ -0,0 +1,91 @@
+---
+name: quickbooks
+description: "QuickBooks Online connector — operator-owned Intuit app, OAuth consent relay, then agent-driven read/write over the QBO REST API."
+tools:
+  - name: quickbooks-credentials-set
+    publicAllowlist: false
+    adminAllowlist: true
+  - name: quickbooks-authorize-url
+    publicAllowlist: false
+    adminAllowlist: true
+  - name: quickbooks-connections
+    publicAllowlist: false
+    adminAllowlist: true
+  - name: quickbooks-token-audit
+    publicAllowlist: false
+    adminAllowlist: true
+  - name: quickbooks-disconnect
+    publicAllowlist: false
+    adminAllowlist: true
+  - name: quickbooks-query
+    publicAllowlist: false
+    adminAllowlist: false
+  - name: quickbooks-report
+    publicAllowlist: false
+    adminAllowlist: false
+  - name: quickbooks-invoice-create
+    publicAllowlist: false
+    adminAllowlist: false
+  - name: quickbooks-invoice-update
+    publicAllowlist: false
+    adminAllowlist: false
+  - name: quickbooks-customer-create
+    publicAllowlist: false
+    adminAllowlist: false
+  - name: quickbooks-customer-update
+    publicAllowlist: false
+    adminAllowlist: false
+  - name: quickbooks-bill-create
+    publicAllowlist: false
+    adminAllowlist: false
+  - name: quickbooks-payment-create
+    publicAllowlist: false
+    adminAllowlist: false
+  - name: quickbooks-entity-create
+    publicAllowlist: false
+    adminAllowlist: false
+  - name: quickbooks-entity-update
+    publicAllowlist: false
+    adminAllowlist: false
+skills:
+  - skills/quickbooks/SKILL.md
+metadata: {"platform":{"optional":true,"pluginKey":"quickbooks"}}
+mcp:
+  command: node
+  args:
+    - ${PLATFORM_ROOT}/lib/mcp-spawn-tee/dist/index.js
+    - ${PLATFORM_ROOT}/plugins/quickbooks/mcp/dist/index.js
+  env:
+    MCP_SPAWN_TEE_NAME: quickbooks
+    LOG_DIR: ${LOG_DIR}
+    PLATFORM_ROOT: ${PLATFORM_ROOT}
+    ACCOUNT_ID: ${ACCOUNT_ID}
+    SESSION_ID: ${SESSION_ID}
+mcp-manifest: auto
+---
+
+# QuickBooks Online
+
+Connects to QuickBooks **Online** — the cloud accounting product (REST/JSON, OAuth 2.0). There is no local company file; the books live in Intuit's cloud, so every call proves app identity plus per-company consent.
+
+SiteDesk ships installs and walks away, so SiteDesk is **never** the OAuth app owner. The operator registers their **own** Intuit app; its client ID/secret is a per-brand secret like the Cloudflare token. Two steps are irreducibly human and are relayed, never auto-clicked: registering the app at developer.intuit.com, and clicking the consent link for each company. Everything after — code→token exchange, access-token refresh, every REST call — is agent-driven.
+
+## Capabilities
+
+- **quickbooks-credentials-set** — store the operator's Intuit app client ID + secret (sandbox or production).
+- **quickbooks-authorize-url** — mint a consent URL for a human to click; one company per click.
+- **quickbooks-connections** — list connected companies and the age of each refresh token.
+- **quickbooks-token-audit** — probe each refresh token; surfaces a dead connection before a customer-facing call hits it.
+- **quickbooks-disconnect** — drop one company's stored token.
+- **quickbooks-query** — read any entity via the QBO query language (invoices, customers, bills, …).
+- **quickbooks-report** — fetch a named report (ProfitAndLoss, BalanceSheet, …).
+- **quickbooks-invoice-create / -update**, **quickbooks-customer-create / -update**, **quickbooks-bill-create**, **quickbooks-payment-create** — entity-specific writes.
+- **quickbooks-entity-create / -update** — generic write for any other entity (Vendor, Item, Estimate, …).
+
+## Setup
+
+Connecting QuickBooks runs through the `quickbooks` skill (`skills/quickbooks/SKILL.md`): register the Intuit app, paste the credentials, then relay the consent link. The skill is the operator-facing procedure; load it and follow its instructions.
+
+## Credentials and isolation
+
+Credentials and per-company refresh tokens live only in this account's secrets directory (`secrets/quickbooks.json`, mode `0600`), inside the brand-isolated install. No Intuit app identity is shared across brands. Token refresh is automatic until a refresh token expires (~100 days of inactivity) or is revoked.

package/payload/platform/plugins/quickbooks/lib/mcp-spawn-tee/index.js

@@ -0,0 +1,193 @@
+#!/usr/bin/env node
+"use strict";
+/**
+ * MCP spawn-tee — in-process stderr capture + lifecycle observability.
+ *
+ * Claude Code spawns each MCP server itself; the platform never holds a
+ * ChildProcess handle. This shim sits between Claude Code and the real MCP
+ * server: Claude Code runs `node <this> <real-entry>`, and the shim runs the
+ * real entry IN ITS OWN PROCESS via dynamic import() — one node runtime, not
+ * two. Before importing, it replaces process.stderr.write with a tee so every
+ * stderr byte the server writes is mirrored to the log sinks; stdin and stdout
+ * (the JSON-RPC channel) are never touched.
+ *
+ *   Claude Code CLI → shim (this file) = node running <real-entry> in-process
+ *        stdin/stdout : untouched (JSON-RPC channel)
+ *        stderr       : process.stderr.write teed → per-date + per-session + passthrough
+ *
+ * Destinations (Task 706) — unchanged:
+ *   - `${LOG_DIR}/mcp-<name>-stderr-<date>.log`  — per-date raw (back-compat;
+ *     the in-process mcp-stderr-tee and the [mcp-init-error] probe read it).
+ *   - `${LOG_DIR}/mcp-<name>-<SESSION_ID>.log`   — per-session raw + lifecycle
+ *     lines. Authoritative sink. Absent when SESSION_ID is unset (enumeration
+ *     spawn) — then the per-date file is the fallback.
+ *   - `server.log` via the loopback log-ingest route — best-effort mirror of
+ *     the [mcp-helper] lifecycle lines.
+ *
+ * Lifecycle lines, correlation key `session=<id8> server=<name>` — unchanged:
+ *   [mcp-helper] op=spawn ... pid= entry=
+ *   [mcp-helper] op=boot  ... head=<first stderr bytes>
+ *   [mcp-helper] op=exit  ... code= signal= lifetimeMs= stderr-tail=
+ *
+ * Process model (Task 989): the shim IS the server's process. op=exit fires
+ * from process.on("exit"), so it covers a normal exit, a non-zero exit, an
+ * uncaught throw (code 1), and a catchable-signal exit (SIGTERM/SIGINT/SIGHUP,
+ * whose handlers record the signal name). An external SIGKILL is uncatchable
+ * and leaves no op=exit line — the per-session stderr tail already on disk and
+ * Claude Code's own transport-drop are the evidence for that death mode.
+ *
+ * The shim never writes to fd 1 (stdout) — that is the JSON-RPC channel.
+ */
+Object.defineProperty(exports, "__esModule", { value: true });
+const node_fs_1 = require("node:fs");
+const node_path_1 = require("node:path");
+const node_url_1 = require("node:url");
+const SERVER_NAME = process.env.MCP_SPAWN_TEE_NAME ?? "unknown";
+const LOG_DIR = process.env.LOG_DIR;
+const SESSION_ID = process.env.SESSION_ID;
+const PLATFORM_PORT = process.env.PLATFORM_PORT;
+const ENTRY = process.argv[2];
+const SESSION_ID8 = SESSION_ID ? SESSION_ID.slice(0, 8) : "—";
+const spawnStamp = Date.now();
+// Per-session raw + lifecycle log (Task 706). Empty SESSION_ID (enumeration
+// spawn) → undefined, and the per-date file is the fallback.
+const perSessionPath = LOG_DIR && SESSION_ID
+    ? (0, node_path_1.resolve)(LOG_DIR, `mcp-${SERVER_NAME}-${SESSION_ID}.log`)
+    : undefined;
+const perDatePath = LOG_DIR
+    ? (0, node_path_1.resolve)(LOG_DIR, `mcp-${SERVER_NAME}-stderr-${new Date(spawnStamp).toISOString().slice(0, 10)}.log`)
+    : undefined;
+// Rolling tail of entry stderr for op=exit. Capped so a chatty server cannot
+// grow the buffer without bound.
+const TAIL_CAP = 2048;
+let stderrTail = "";
+let bootEmitted = false;
+let exitEmitted = false;
+let exitSignal;
+// The real stderr writer, captured before the tee replaces it. Lifecycle lines
+// and stderr passthrough both go through this so they are never re-teed into
+// the per-date sink nor recursed back into the patched writer.
+const rawStderrWrite = process.stderr.write.bind(process.stderr);
+// LOG_DIR is created once, lazily, on the first successful append — not per
+// chunk. teeWrite runs on the server's own stderr hot path now, so a per-call
+// mkdirSync would be two syscalls per log line for nothing.
+let logDirReady = false;
+function appendSafe(path, data) {
+    if (!path || !LOG_DIR)
+        return;
+    try {
+        if (!logDirReady) {
+            (0, node_fs_1.mkdirSync)(LOG_DIR, { recursive: true });
+            logDirReady = true;
+        }
+        (0, node_fs_1.appendFileSync)(path, data);
+    }
+    catch {
+        /* unwritable destination — never mask primary output */
+    }
+}
+// Best-effort mirror of a lifecycle line to server.log via the loopback
+// log-ingest route. Fire-and-forget: the per-session file is authoritative, so
+// a dropped POST loses nothing. On the "exit" event the loop is stopping, so
+// the op=exit mirror may not flush — the per-session line, written sync, holds.
+function postToServerLog(suffix, level) {
+    if (!PLATFORM_PORT)
+        return;
+    try {
+        const ctrl = new AbortController();
+        const t = setTimeout(() => ctrl.abort(), 500);
+        t.unref?.(); // never let the abort timer keep the shim's event loop alive
+        void fetch(`http://127.0.0.1:${PLATFORM_PORT}/api/admin/log-ingest`, {
+            method: "POST",
+            headers: { "content-type": "application/json" },
+            body: JSON.stringify({ tag: "mcp-helper", level, line: suffix }),
+            signal: ctrl.signal,
+        }).then(() => clearTimeout(t)).catch(() => clearTimeout(t));
+    }
+    catch {
+        /* fetch threw synchronously — best-effort only */
+    }
+}
+// Emit one [mcp-helper] lifecycle line: authoritative per-session file (sync,
+// survives process.exit), the shim's own stderr via the RAW writer (journald /
+// Claude Code visibility, never re-teed), and a best-effort server.log mirror.
+// `suffix` is the line body after the tag and carries no newline (the
+// log-ingest route rejects newlines).
+function emitLifecycle(suffix, level) {
+    const line = `[mcp-helper] ${suffix}\n`;
+    appendSafe(perSessionPath, line);
+    try {
+        rawStderrWrite(line);
+    }
+    catch { /* stderr closed */ }
+    postToServerLog(suffix, level);
+}
+if (!ENTRY) {
+    emitLifecycle(`op=error session=${SESSION_ID8} server=${SERVER_NAME} reason="no entry given (argv[2] missing)"`, "error");
+    process.exit(2);
+}
+// Replace process.stderr.write with a tee: mirror every server stderr byte to
+// the per-date and per-session sinks, keep a rolling tail for op=exit, emit
+// op=boot on the first bytes, then pass the write through to the real stderr.
+const teeWrite = ((...args) => {
+    const chunk = args[0];
+    appendSafe(perDatePath, chunk); // per-date raw (back-compat)
+    appendSafe(perSessionPath, chunk); // per-session raw (Task 706)
+    const text = typeof chunk === "string" ? chunk : Buffer.from(chunk).toString("utf8");
+    stderrTail = (stderrTail + text).slice(-TAIL_CAP);
+    if (!bootEmitted) {
+        bootEmitted = true;
+        const head = text.split("\n")[0].slice(0, 200);
+        emitLifecycle(`op=boot session=${SESSION_ID8} server=${SERVER_NAME} head=${JSON.stringify(head)}`, "info");
+    }
+    return rawStderrWrite(...args); // passthrough
+});
+process.stderr.write = teeWrite;
+// Catchable-signal handling. The shim drives the exit ONLY when it is the sole
+// listener for the signal — i.e. the imported entry installed no handler of its
+// own. In that case it records the signal (so op=exit carries signal=<sig>) and
+// exits with 128+signum, mirroring the prior wrapper's exit-status convention.
+//
+// When the entry DID install a handler, the shim defers entirely and records
+// nothing: the entry's handler decides the exit code, and op=exit reflects that
+// exit verbatim. This matches the prior two-process model, where a child that
+// caught the signal and exited cleanly was observed as code=0 signal=— (a clean
+// self-exit), not as a signal death. Setting exitSignal here unconditionally
+// would mislabel every graceful signal-driven shutdown as a signal kill.
+function onSignal(sig, code) {
+    return () => {
+        if (process.listenerCount(sig) === 1) {
+            exitSignal = sig;
+            process.exit(code);
+        }
+    };
+}
+process.on("SIGTERM", onSignal("SIGTERM", 143));
+process.on("SIGINT", onSignal("SIGINT", 130));
+process.on("SIGHUP", onSignal("SIGHUP", 129));
+// op=exit, emitted once from the process "exit" event. Sync-only — the loop is
+// stopping. When a catchable signal caused the exit, report code=— signal=<sig>
+// to match the prior wrapper's format; otherwise code=<exit code> signal=—.
+process.on("exit", (code) => {
+    if (exitEmitted)
+        return;
+    exitEmitted = true;
+    const lifetimeMs = Date.now() - spawnStamp;
+    const tail = stderrTail.slice(-200);
+    const codeField = exitSignal ? "—" : String(code);
+    const level = exitSignal || code !== 0 ? "error" : "info";
+    emitLifecycle(`op=exit session=${SESSION_ID8} server=${SERVER_NAME} pid=${process.pid} ` +
+        `code=${codeField} signal=${exitSignal ?? "—"} lifetimeMs=${lifetimeMs} stderr-tail=${JSON.stringify(tail)}`, level);
+});
+emitLifecycle(`op=spawn session=${SESSION_ID8} server=${SERVER_NAME} pid=${process.pid} entry=${ENTRY}`, "info");
+// Run the real MCP server in THIS process. Dynamic import() (not top-level
+// await — this file compiles to CommonJS) loads the ESM entry; a load-time
+// failure mirrors the old child spawn-error path (op=error, exit 127). The
+// op=exit handler is suppressed on this path so op=error stays the sole record.
+import((0, node_url_1.pathToFileURL)(ENTRY).href).catch((err) => {
+    const msg = err instanceof Error ? err.message : String(err);
+    exitEmitted = true;
+    emitLifecycle(`op=error session=${SESSION_ID8} server=${SERVER_NAME} reason=${JSON.stringify(`import error: ${msg}`)}`, "error");
+    process.exit(127);
+});
+//# sourceMappingURL=index.js.map

package/payload/platform/plugins/quickbooks/lib/mcp-spawn-tee/package.json

@@ -0,0 +1,3 @@
+{
+  "type": "commonjs"
+}