antigravity-operator-core

name	antigravity-operator-core
description	Repo-local Antigravity operator plugin skill. Use when the user asks to use Antigravity Operator Core, route messy context, show a Virtuoso Trace, start Autopilot, create Mission Mode, audit plugin readiness, inspect system health, review routing intelligence, run Extraction Governor, use Knowledge Librarian, or decide whether to self-evolve a workflow.

Antigravity Operator Core

Use this repo-local plugin skill as the hot operating layer for /Users/farricecain/Codex Antigravity.

Route Order

1. Raw context, unclear route, or "what should I do with this?" -> .agent/workflows/autopilot.md
2. Long-running, system-changing, reusable, or multi-milestone work -> .agent/workflows/mission.md
3. User asks for a menu or candidate routes -> .agent/workflows/orchestrate.md
4. Source-to-system, skill boundary, extraction, or forge decision -> .agent/workflows/extraction-governor-agent.md
5. Library overlap, dormant knowledge, or reusable solution check -> .agent/workflows/knowledge-librarian.md
6. System health, dormant loops, or proof-of-life -> .agent/workflows/health-check.md
7. Route quality, misroutes, or usage evidence -> .agent/workflows/routing-intelligence.md
8. Repeated failure, plateau, or workflow improvement -> .agent/workflows/self-evolve.md
9. Full-arsenal, subagent, composition, plugin/tool-blending, agent-elevation, solo orchestration, or full-system excellence work -> .agent/workflows/virtuoso.md
10. Deep research, wide research, social listening, market intelligence, PMF/OMF, ICP deep canvassing, or claim verification -> .agent/workflows/deep-research-os.md
11. Plugin packaging decision -> .agent/workflows/plugin-readiness-audit.md

Virtuoso Trace

For complex work, run:

python3 execution/virtuoso_orchestration.py "[goal]" --json

Preferred command surface:

/virtuoso [goal]

Preferred deep-research surface:

/virtuoso --mode research [research objective]

Route into .agent/workflows/deep-research-os.md when the work needs source-backed research planning, social listening, wide decomposition, source ledgers, claim labels, and anti-hallucination verification.

Use --delegate-intent when the user explicitly asks for subagents, delegated agents, parallel agents, or swarm work. This prepares subagent-first packets but does not spawn real Codex subagents. Use --log-routing only after the chosen compound stack is actually used, so routing intelligence records real ensemble evidence.

The trace must expose:

• primary route and function owner
• recommended stack or skip reason
• bounded composition slots
• delegation matrix and receipt boundary
• plugin/tool surface
• routing evidence
• verifier plan and first action
• execution receipt that separates considered gates from executed workflows/scripts

Rules

• Prefer the current workspace workflow files when they exist.
• Do not replace Antigravity routers; use them.
• Ask before external writes, paid tools, destructive changes, or real subagents.
• Keep system-audit, expert-composition-governor, and source-to-skill-system as referenced dependencies until plugin readiness scores justify bundling them.
• For plugin packaging, run python3 execution/plugin_readiness_audit.py before creating or expanding a plugin.
• A plugin is proven only after direct invocation, natural language trigger, missing-info behavior, path resolution, and fresh-session checks pass.