<domain Slug>/<sub System Slug>

Skill (inside )

Purpose

<One paragraph. What this skill does. What instrument it carries. What research grounds it. What artifacts it produces. Not a restatement of the agent's Identity — the skill is the *how*, the agent is the *who*.>

Activation

This skill fires when:

• Any /<SLUG>-* command is invoked
• Keywords match: <keyword list>
• The starlight-<SLUG> agent is active
• <other activation signals>

The skill does NOT fire when:

• The request is substrate-level (use Starlight's SKILL.md)
• The request belongs to a sister sub-system (defer to sister)
• The request is outside <DOMAIN> entirely

Protocol

This is the 7-step protocol this skill runs. Matches the agent's Reasoning Protocol but framed as skill-level operational steps.

Step 1 — <Phase name>

Required input: <input> Output: <output> Halt condition: <when this step refuses to proceed>

Step 2 — <Phase name>

<...>

Required input: <input> Output: <output> Halt condition: <...>

Step 3 — <Phase name>

<...>

Step 4 — <Phase name>

<...>

Step 5 — <Phase name>

<...>

Step 6 — <Phase name>

<...>

Step 7 — Ship artifact + attest

Every artifact ships with:

• <DOMAIN>-specific disclaimer block at the top (legal / clinical / jurisdictional)
• Research citations where claims are made (direction, not invented numbers)
• Named hand-off (one, not a menu)
• "Built on SIP" attestation footer

Path convention: verticals/<vertical-slug>/<sub-system-slug>/<artifact-type>-<context>-<YYYY-MM-DD>.md

Rules

1. <domain-specific rule 1 — e.g., "Never ship without disclaimer when jurisdictional territory is touched">
2. <domain-specific rule 2>
3. <domain-specific rule 3>
4. Research grounding required where claims are made. Cite direction of evidence. Do not invent effect sizes or specific percentages.
5. One hand-off at close. No menus. Named command or named real-world move.
6. Built on SIP attestation on every artifact. Sovereignty clause non-waivable.

Composition

This skill composes with:

• <sister sub-system skill> — <how>
• intelligence/decision-framework — <when option-collapse is load-bearing>
• intelligence/pattern-recognition — <when longitudinal / multi-signal pattern reading is load-bearing>
• memory/knowledge-synthesis — <when composing artifact from multiple protocol-step outputs>

Does NOT compose with:

• <skills that would drift this one out of scope>

Built on SIP — domain-stack-starter sub-system skill template · v7.4.1 · SIP v1.1.0