Sync clarion-setup: leaner install + free-tier reasoning default

[jingerzz]

Summary

Syncs two changes from jingerzz/clarion-intelligence-system into the registry mirror at External/clarion-setup/:

1. SKILL.md — persona/rule install removed from autonomous path. The clarion-setup skill no longer installs the 7 Clarion personas + 8 routing rules during its autonomous install. Those moved to an opt-in follow-up prompt described in docs/PERSONAS-AND-RULES.md. Skills are functional standalone; the persona layer is chat-UX discipline.

2. setup.py — sync free-tier reasoning_model default. Carries jingerzz/clarion-intelligence-system#23 into the registry, which was never back-synced. Bootstrap default changes from subscriber-tier zo:anthropic/claude-opus-4-7 to free-tier zo:openai/gpt-5.4-mini, matching the upstream "all defaults are free-tier" policy.

Motivation

• Install speed (change 1). A real user install transcript showed the autonomous portion of clarion-setup took ~7 minutes — of which ~5 minutes was the agent parsing the 62KB PERSONAS-AND-RULES.md doc (nested triple-backtick code fences in persona prompts are agent-hostile) and calling create_persona × 7 + create_rule × 8 sequentially. Moving personas/rules out of the autonomous path drops the install to ~30 seconds. Users who want persona routing opt in afterward via a separate prompt; the bare install still gives them the working library, all sibling skills, and the registered sec-indexer service.

• Free-tier consistency (change 2). Fresh installs from the registry today get zo:anthropic/claude-opus-4-7 as the reasoning_model default — a subscriber-tier model. Free-tier users hit tier-access errors. The upstream policy in ARCHITECTURE.md's "Default model selection" is "all defaults are free-tier"; this PR aligns the registry-shipped setup.py with that policy.

Diff summary

File	Change
External/clarion-setup/SKILL.md	Removes Steps 4 (personas) + 5 (rules). Renumbers remaining steps (4 = human checkpoint, 5 = restart, 6 = verify, 7 = report). Updates frontmatter description, Flow at a glance intro, idempotency + on-error sections.
External/clarion-setup/scripts/setup.py	One-field change: DEFAULT_CONFIG["reasoning_model"] from zo:anthropic/claude-opus-4-7 → zo:openai/gpt-5.4-mini. Adds an explanatory comment.

Validation

$ bun validate

Passes. Warnings exist for other External skills (theme-factory, xlsx, etc.) but none for External/clarion-setup.

Test plan

• After merge, a fresh install the clarion-setup skill invocation from chat should complete the autonomous portion in ~30 seconds (vs ~7 minutes today)
• Service registration + restart + service_doctor verification all happen automatically as before
• No persona/rule creation happens during install (verifiable by running list_personas() and list_rules() post-install — counts unchanged from pre-install)
• Existing users who already have Clarion personas/rules from a previous install retain them untouched (re-runs of clarion-setup don't modify Zo Settings)

Upstream context

Upstream PR (carries the same SKILL.md change + companion doc updates in README.md, ARCHITECTURE.md, docs/TEST-PLAN.md, docs/PERSONAS-AND-RULES.md): jingerzz/clarion-intelligence-system#27

🤖 Generated with Claude Code