Team Agents Orchestrator#

Use this skill when one task spans multiple modes (research + implementation + communication) and quality can degrade without structure.

Roles#

• Operator: owns user interaction and final decisions
• Researcher: gathers facts/options/constraints
• Builder: executes implementation steps and reports exact diffs/errors
• Editor: refines final messaging for clarity and actionability

Activation criteria#

Activate when any of these are true:

• task is multi-step and high impact
• task crosses domains (e.g., web research + config changes + user-facing output)
• task has external/public consequences
• task likely needs iterative review

Orchestration loop#

1. Define objective and constraints.
2. Create a handoff contract (see references/handoff-contract.md).
3. Run role steps in order (Researcher → Builder → Editor).
4. Operator reviews and decides final output/action.
5. Log durable outcomes and follow-ups.

Handoff discipline#

Every handoff must include:

• objective
• constraints
• output format
• done criteria
• open questions

Receiver must restate scope before executing.

Memory hygiene rules#

• Durable outcomes/decisions → daily memory
• Stable preferences/facts → long-term memory
• Temporary chatter/debug text should not be promoted unless reused

Common failure modes#

• role blur (single messy pass)
• overreach without approval
• verbose output without added value
• memory pollution from transient logs

Output contract#

Return:

• concise final result
• what changed
• unresolved risks/questions
• next best action