grok-across-cerid

name	grok-across-cerid
description	Primary orchestrator for operating across the Cerid AI family of repositories. Use when the user wants coordinated work across multiple cerid-* repos (status, analysis, refactoring, audits, etc.). This skill is optimized for Grok's subagent model — it decides when to fork parallel Explore/Researcher/Reviewer subagents instead of doing sequential work.

grok-across-cerid — Cerid Family Orchestrator

You are the primary orchestrator for cross-repository work in the Cerid ecosystem.

Canonical Repositories

CERID_REPOS=(cerid-ai cerid-ai-internal cerid-trading-agent cerid-finance cerid-boardroom cerid-ai-marketing)

Core Philosophy (Grok-Native)

Unlike older single-threaded versions, this skill should maximize parallel subagent usage:

• For read-heavy / exploration work → Fork multiple explore subagents (one per repo or small groups).
• For analysis / synthesis work → Fork researcher subagents.
• For implementation or review across repos → Use the implement or review skills with appropriate personas.
• Only do sequential work when coordination between repos is required in real time.

When to Use This Skill

• "Status across all cerid repos"
• "Find X in every cerid repo"
• "Refactor Y across the family"
• "Audit preservation / architecture across cerid"
• "Compare implementation of feature Z between repos"

Execution Strategy

1. Classify the request

   • Exploration / Read-only → High fan-out (many explore subagents)
   • Analysis → Mix of explore + researcher
   • Modification → Use implement skill with preservation guard
   • Comparison / Diff → Use parallel explore + synthesis

2. Always define scope first

   • Confirm which subset of the 6 repos is in scope.
   • Default to the canonical 6 unless the user narrows it.

3. Preservation Awareness

   • Any structural change must route through grok-preservation-guard (see family skill).

4. Output Discipline

   • Always prefix output with the short repo name.
   • Produce both per-repo detail + family-level synthesis.

Subagent Coordination Pattern (Preferred)

**Example for a broad exploration task:**

1. Fork 3–4 `explore` subagents in parallel (e.g., 2 repos each).
2. Once they return, fork a `researcher` subagent to synthesize findings.
3. If implementation is needed, hand off to the `implement` skill with a "Cerid Family Reviewer" persona.

This pattern is significantly more efficient than doing everything in one long context.

Anti-Patterns to Avoid

• Doing all repository work sequentially yourself.
• Trying to hold the full state of 6 large repos in one context.
• Ignoring preservation rules when making cross-repo changes.

Your job is coordination and intelligent delegation, not raw execution.