Skills export

SKILLS.md — the agent interface.

BrainOS distills ingested knowledge into atomic, agent-readable instructions. Pick a department to give an agent only what its function needs — a Legal agent doesn't need finance ledgers, a Finance agent doesn't need on-call runbooks.

No knowledge to export yet. Ingest something to generate the skills file.

How to load this into an agent

Claude CodeCLAUDE.md

Drop the chosen department's SKILLS.md into your project root as CLAUDE.md. Every Claude Code session starts pre-loaded.

# Engineering team:
curl /api/skills?department=engineering > CLAUDE.md

Anthropic SDKsystem prompt

Inject at agent startup. Pick the department matching the agent's role — Legal agent gets Legal+General, no Finance.

skills = open("SKILLS-legal.md").read()
client.messages.create(
  system=skills,
  messages=[...]
)

Multi-agent systemrole-aware

Each specialized agent loads only its department's skills file. Smaller context, lower cost, no leakage between functions.

agents = {
  "legal":   load("?department=legal"),
  "finance": load("?department=finance"),
  "eng":     load("?department=engineering"),
}

API endpoints

GET /api/skills→ full SKILLS.md (all departments)

GET /api/skills?department=legal→ Legal + General only

GET /api/skills?format=json→ structured JSON for programmatic use

GET /api/skills?department=hr&format=json→ HR-only JSON