Commands

# /request-skill — Report a Missing Skill

Use this when you know a skill is missing but don't want to build it yourself.
Takes 4 questions, checks existing coverage, and adds a TODO row to `MASTER_QUEUE.md`.

---

## Step 1 — Ask the 4 questions

Ask the user exactly these questions. Ask all 4 upfront, not one at a time:

```
1. What Salesforce task does this skill cover?
   (Describe it as a task someone performs, not a feature name)
   Example: "Configure case assignment rules in Service Cloud"
   Not: "Case assignment rules"

2. Who is this skill for?
   Admin / BA / Developer / Data / Architect

3. Which Salesforce cloud does this apply to?
   Core Platform / Sales Cloud / Service Cloud / Experience Cloud /
   Marketing Cloud / Revenue Cloud / Field Service / Health Cloud /
   Financial Services Cloud / Nonprofit Cloud / Commerce Cloud /
   Agentforce / OmniStudio / CRM Analytics / Integration / DevOps

4. What's the main problem it solves or mistake it prevents?
   (One sentence)
```

---

## Step 2 — Check for existing coverage

```bash
cd /path/to/SfSkills
python3 scripts/search_knowledge.py "<task from question 1>"
python3 scripts/search_knowledge.py "<task from question 1>" --domain <domain>
```

**If `has_coverage: true`:**
Show the user the existing skill. Ask: "Does this cover what you need, or is your use case different enough to warrant a new skill?"
- If covered → stop. Point them to the existing skill.
- If different → continue with a clear disambiguation note.

**If `has_coverage: false`:**
Continue.

---

## Step 3 — Determine domain and skill name

Map role → domain:
```
Admin / BA / Architect role → admin
Developer role → most specific: apex / lwc / flow / integration / devops
Data role → data
OmniStudio topics → omnistudio
Agentforce topics → agentforce
Security topics → security
```

Generate a kebab-case skill name from the task:
- "Configure case assignment rules" → `case-assignment-rules`
- "Set up territory management for Sales Cloud" → `sales-cloud-territory-management`
- "Apex callout error handling patterns" → `callout-error-handling`

Rules:
- All lowercase, hyphens only
- No cloud prefix unless the skill is cloud-specific and ambiguous without it
- Keep it short — 2-4 words is ideal

---

## Step 4 — Write the description

Compose a one-sentence description:

```
"Use when [trigger scenario — what the practitioner is doing when they need this].
Triggers: [2-3 keyword phrases]. NOT for [explicit exclusion]."
```

Example:
```
"Use when configuring automatic routing of new cases to queues or users.
Triggers: 'set up case routing', 'auto-assign cases', 'case queue assignment'.
NOT for opportunity routing or lead assignment rules."
```

The "NOT for..." is mandatory. Without it, the skill will be rejected by the Validator.

---

## Step 5 — Add to MASTER_QUEUE.md

Find the correct section in `MASTER_QUEUE.md`:
- Phase 1 = Core Platform
- Phase 2 = Sales Cloud
- Phase 3 = Service Cloud
- (etc. — match the cloud from question 3)

Find the table for the correct role within that phase.

Add the row:
```markdown
| TODO | <skill-name> | <description from Step 4> | Requested by user |
```

If the phase section doesn't exist yet (the cloud hasn't been researched):
Add to the relevant Research table instead:
```markdown
| TODO | <skill-name> | <description> | Requested — skip research gate for this skill |
```

---

## Step 6 — Confirm with the user

Report back:
```
✓ Skill request added to queue: <domain>/<skill-name>
  Phase: <phase number> — <cloud name>
  Role: <role>
  Queue position: approximately #N in the build queue

The Orchestrator will pick this up on the next run.
Track it in MASTER_QUEUE.md under [Phase] → [Role] table.
```

---

## What This Command Does NOT Do

- It does not build the skill. Building goes through Content Researcher → Role Agent → Validator.
- It does not guarantee a build timeline — the queue runs at its own pace.
- It does not accept requests for skills that already exist — check first, always.