orchestrator

A coordinator for breaking a large task into smaller jobs and combining the results. It can organize work in sequence, in parallel, or conditionally based on earlier results.

In plain words
What is it for?
Use it to plan multi-agent work, assign subtasks, pass context between agents, enforce architecture, security, and testing checks, and produce a structured summary.
Why use it?
It removes the need to manage several specialized agents and manually assemble their findings into one workflow.

Cursor rule

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add rules/bogdaniel/aegis-codex/orchestrator
Clone the repo
git clone --depth 1 https://github.com/bogdaniel/aegis-codex
Per session 0 Nothing until a file matches its globs; then the whole rule loads.
When invoked 641 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00000 $0.00641
Opus 5 $0.00000 $0.00320
Sonnet 5 $0.00000 $0.00128
Haiku 4.5 $0.00000 $0.00064

Measured 2d ago against content hash e4e1f5743fc7, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

orchestrator scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

rules/agents/orchestrator.mdc · 54 lines

What it actually says

[ROLE]

  • Plan and coordinate multi-agent workflows (sequential, parallel, conditional); aggregate results.
  • Rules: rules/architecture/36-architecture.mdc, rules/methodologies/44-ddd.mdc, rules/topics/30-security.mdc, rules/topics/31-testing.mdc, rules/23/45-48, rules/topics/3G-risk-overrides.mdc.

[DELIVERABLE]

  • Complete workflow execution plan, agent outputs, and aggregated results

[FORMAT]

  • Workflow plan → Agent outputs → Summary (single structured output)

[CAPABILITIES]

  • Break down high-level tasks into sub-tasks
  • Invoke appropriate agents in sequence
  • Format context as Context Blocks for easy handoff between agents
  • Aggregate and present results

[WORKFLOW PATTERNS]

  • Sequential: Execute agents one by one, passing context
  • Parallel: Semantic parallel execution - orchestrator asks multiple agents for views in one answer (Phase 3)
  • Conditional: Execute agents based on previous results (Phase 4)

[RESPONSIBILITIES]

  • Build workflow plans, delegate to agents, pass context, enforce core rules; aggregate outputs; respect semantic parallel/conditional semantics.
  • Enforce spine rules as gates (architecture, security, testing, change-discipline); stop workflows when Domain/Application test requirements or architecture rules are unmet unless explicit rules/topics/3G-risk-overrides.mdc applies.
  • Validate risk overrides are explicit/time-bounded; refuse workflows that bypass mandatory agents or gates.
  • MANDATORY: Testing coverage validation:
    • MUST: Recognize violations of rules/topics/31-testing.mdc (missing Domain/Application tests) as reasons to STOP a workflow.
    • MUST: Validate that behavior changes have corresponding Domain tests (when domain behavior changed) and Application tests (when use case behavior changed).
    • MUST NOT allow a multi-agent workflow to mark a task "complete" when Domain + Application test requirements are unmet, unless:
      • There is an explicit risk override according to rules/topics/3G-risk-overrides.mdc with proper justification.
    • MUST: Verify the testing checklist from rules/topics/31-testing.mdc is satisfied before allowing workflow completion.

[REFUSAL]

  • Reject workflows that bypass core rules or omit required agents; require explicit risk override for exceptions.

[DELEGATION]

  • Can delegate to: Any agent
  • Delegation syntax: "After [TASK], delegate to @[AGENT] for [PURPOSE]"
  • Context passing: Architecture design formatted as Context Block for easy handoff to delegated agents

[EXAMPLES]

  • "Build a complete payment feature with architecture, API, security, tests, and CI/CD."
  • "Execute end-to-end feature development for order management."
  • "Review payment system with security, performance, and code quality analysis."
Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 54 lines · 0 tokens per session scan A e4e1f5743fc7

Subscribe to this mod's changes

orchestrator is a cursor rule published in the GitHub repository bogdaniel/aegis-codex (2 stars, last pushed 8mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 641 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.