t0ddc3by/claude-for-customer-success/cold-start-interview

/cold-start-interview

[PROPOSED]

Use When

• Running the CSM plugin for the first time on a new installation
• The CSm practice has changed materially (new segments, motion change, new CS platform)
• Config files are missing or contain [PLACEHOLDER] markers that block other skills
• The user explicitly invokes /csm:cold-start-interview or /csm:customize --full

Do NOT Use For

• Updating a single config section — use /csm:customize --section <name> instead
• Answering ad-hoc CS strategy questions — answer directly without running the interview
• Reconfiguring other plugins (cs-ops, renewals, onboarding) — each has its own customize skill

Typical Activation

"Set up the CSM plugin for my company" "Run the cold start interview" "/csm:cold-start-interview" "Configure the plugin from scratch"

1. Check ~/.claude/plugins/config/claude-for-customer-success/csm/CLAUDE.md. If --check-integrations, skip the interview — re-run only the Part 0 What's connected? probe and rewrite the ## Available integrations table at that config path. When probing: only report ✓ if an MCP tool call actually succeeded. Configured-but-untested connectors are marked ⚪ with a one-line how-to for confirming. Never report ✓ based on .mcp.json declarations alone.
2. Check the shared company profile at ~/.claude/plugins/config/claude-for-customer-success/company-profile.md. If it exists, read it and skip company-level questions. If it doesn't exist, write it after the interview.
3. Run the interview below: Part 0 (role + integrations), then CS motion, portfolio, health model, escalation, and playbook.
4. Write ~/.claude/plugins/config/claude-for-customer-success/csm/CLAUDE.md, creating parent directories as needed.

Cold-Start Interview: Customer Success Manager

Pre-flight

Check for existing config files at:

• ~/.claude/plugins/config/claude-for-customer-success/csm/CLAUDE.md
• ~/.claude/plugins/config/claude-for-customer-success/company-profile.md

For --full or no argument: proceed regardless of whether files exist — this skill creates or overwrites them. For --section <name>: load existing config file content before displaying current values alongside each question. For --reset: display a destructive-clear warning and require explicit confirmation before proceeding.

G-code dependency: All G-code guardrails referenced in this skill (G1–G9) are defined in the CLAUDE.md config loaded above. If Pre-flight halts or config is missing, G-codes are undefined — do not proceed with partial config.

Reasoning Protocol

Before generating output, apply these primers:

1. CLASSIFY: What type of cold-start setup is this?

• Type A — Fresh Install: No config file exists. No prior profile. Full or quick interview needed.
• Type B — Resume from Pause: Config exists with <!-- SETUP PAUSED AT: --> marker. Partial answers already written.
• Type C — Redo (Full or Section): Populated config exists. User invoked --redo or --redo <section>.
• Type D — Integration Check Only: User invoked --check-integrations. No interview — probe connectors and rewrite the integrations table.
• Type E — Quick Path: User chose the 2-minute setup. Only Part 0 + CS motion. Everything else gets motion-aware [DEFAULT] markers. Outcome catalog gets [PENDING] — surface the --generate-outcome-catalog flag in the Close section.
• Type F — Catalog Generation Only: User invoked --generate-outcome-catalog. Skip the full interview. Read company-profile.md to retrieve product name and primary value metric, then run Part 6 only.

1. CONSTRAINTS: What limits the solution space?

• Is the working directory project-scoped (limited filesystem access) or user-scoped?
• Does a shared company profile already exist (skip company questions) or not (ask and write it)?
• Which connectors are available — test with actual calls, not config declarations?
• What interview path did the user choose — quick or full?

1. EXPERT CHECK: What would a veteran CSM verify first?

• Are there existing artifacts (playbooks, escalation docs, CRM exports) that replace manual Q&A?
• Does the CS motion answer (high-touch / tech-touch / hybrid / handoff) correctly drive all downstream defaults?
• On resume or redo: are previously written answers still accurate before continuing?
• On integration check: is the connector actually responding, or just configured?

1. ANTI-PATTERNS: Common mistakes to avoid:

• ❌ Asking more than 2-3 answerable prompts per turn (count subparts as separate questions)
• ❌ Reporting a connector as connected based on .mcp.json without an actual successful tool call
• ❌ Writing generic defaults without [DEFAULT] markers, making them indistinguishable from user-provided values
• ❌ Re-asking company-level questions when the shared company profile already exists
• ❌ Overwriting a populated config section during a targeted --redo <section> without preserving adjacent sections

After execution, verify:

• Does the written config satisfy both the user's explicit answers AND motion-appropriate defaults for unanswered fields?
• Were all connector statuses verified by actual tool calls, not config inspection?
• Confidence: [High] if full interview completed with user-provided answers; [Medium] if quick path with motion-aware defaults; [Low] if resumed from stale pause with unconfirmed prior answers

Purpose

Every CSM skill reads from the configuration this interview writes. A generic profile gives generic output — a default health model, a default escalation matrix, a default set of risk flags that may have nothing to do with how your team actually works. This interview builds a profile calibrated to your accounts, your motion, and your risk posture.

Cold-start check

Read ~/.claude/plugins/config/claude-for-customer-success/csm/CLAUDE.md:

• Does not exist → start the interview.
• Contains `<!-- SETUP PAUSED AT: -->` → greet the user and offer to resume from that section.
• Contains `[PLACEHOLDER]` markers but no pause comment → the template was never completed; offer to start fresh or resume from wherever the placeholders begin.
• Populated (no placeholders, no pause comment) → already configured; skip unless --redo.

The template structure lives at ${CLAUDE_PLUGIN_ROOT}/CLAUDE.md — use it as the section scaffold. Write the completed company profile to the config path, creating parent directories as needed.

Check for the shared company profile

Look for ~/.claude/plugins/config/claude-for-customer-success/company-profile.md.

• If it exists: Read it. Show a one-line confirmation: "You're at [company], [product description], in the [segment] segment. Right? (Or say 'update' to change the shared profile.)" If confirmed, skip company questions — go straight to CSM-specific ones.
• If it doesn't exist: This is the first plugin install. After the fork choice, ask the company questions (company name, product description, primary segment, CS team size, churn drivers) and write them to the shared profile. Tell the user: "I've saved your company profile — the other CS plugins will read it and skip these questions."

Company-level questions that live in the shared profile and should NOT be re-asked: company name, product description, primary segment, CS team size, top churn drivers. CSM-specific questions (health model config, account portfolio, escalation matrix, playbook sources) stay in the per-plugin config.

Install scope check

Before the orientation, if the working directory is inside a project (not the user's home directory), flag it once:

Heads up — it looks like this plugin may be project-scoped, which means I can only read files in [current directory]. If you'll want me to pull documents from elsewhere (success plans, QBR decks, account notes in other folders), install user-scoped instead — see QUICKSTART.md. You can continue with project scope, but you'll need to move files into this folder.

Ask the user to confirm before proceeding. If the working directory is the user's home directory, skip this check silently.

Before the interview starts

Open with the fork-first preamble:

`csm` is for CSMs managing a B2B SaaS book of business — account research, call prep, QBRs, success plans, health reviews, risk flags, and renewal readiness. Not your role? /claude-for-customer-success:related-plugins.

2 minutes gets you your role, CS motion, and tool connections — enough to run most skills with sensible defaults. 15 minutes builds your full profile: account portfolio details, health model configuration, escalation matrix, and playbook sources — the things that turn generic output into output that reads like you wrote it.

Quick or full?

Quick path: Ask only Part 0 and the CS motion section. Write the config with [DEFAULT] markers on everything else. Close with: "Done. You can start using the skills now. I've used sensible defaults for health thresholds, risk signals, and escalation routing. When a skill's output feels off, that's usually a default to tune — the output will tell you which one. Run /csm:cold-start-interview --full anytime to do the whole interview."

Full path: The existing interview flow below. After the user picks, give the orientation, then proceed.

After the user picks quick or full

Give the orientation, in your own voice:

"This plugin maintains your CSM company profile: your CS motion, your account portfolio context, your health model, and your escalation chain. It reads from a plain-text config file every time a skill runs. Everything you answer can be changed later — either by editing the file directly or re-running this interview."

Then:

"Setup builds a fresh professional profile from your answers. It doesn't read your Claude history, other conversations, or personal CLAUDE.md files. If I notice relevant context already in our conversation, I'll ask before using it."

Then: "Ready? A few quick questions first."

Interview pacing

• Assume the answer exists somewhere. For anything that's probably written down — CS playbook, escalation matrix, health model config — prompt for a paste or link before asking the user to type it from memory. "Paste a doc, share a link, or give me the short version" is the default ask for any question that's more than a sentence.
• Batch size. Never ask more than 2-3 answerable prompts per turn. Count subparts. One question with 5 subparts is 5 questions.
• Pause and resume. Tell the user up front: "Say 'pause' anytime and I'll save progress. Run /csm:cold-start-interview again to pick up where you left off." When the user pauses, write a partial config with <!-- SETUP PAUSED AT: [section] — run /csm:cold-start-interview to resume --> at the top and [PENDING] markers (distinct from [PLACEHOLDER]) on unanswered fields.

The interview

Part 0: Who's using this, and what's connected

Three quick orientation questions before CS specifics. These shape how every skill frames its output.

Who's using this?

Who'll be using this plugin day to day?

1. CSM — you own the full post-sales relationship: onboarding, adoption, QBRs, renewal conversations, expansion signals.

2. CS team lead or manager — you manage CSMs and use this for portfolio reviews, escalation decisions, and coaching.

3. CS adjacent — Solutions Engineer, Onboarding Specialist, or Renewals Manager who occasionally needs CSM-type output.

This calibrates output depth and escalation framing across every skill.

Your CS motion