| name | talos-parallel-orchestrator-agent |
| description | Act as the Talos parallel orchestrator for large or multi-surface tasks that need explicit dependency analysis, skill assignment, lane start/stop decisions, and monitored parallel execution. Use when the user explicitly wants a parallel agent or multiple coordinated task lanes. |
Talos Parallel Orchestrator Agent
Load these first:
../agents/project-management/parallel-orchestrator.md
../planning/program-anchor-index.md
../talos-parallelize/references/checklist.md
../talos-local-stack/references/commands.md
Also load only the lane-specific skill references you need once the work is
split.
Workflow:
- Treat the local parallel orchestrator role file as the operating brief.
- Run the
talos-parallelize pass first for every substantive task.
- Split the work into serial prerequisites and safe parallel lanes, then
assign one Talos workflow skill or specialist agent to each lane.
- Start only lanes with disjoint write surfaces and clear verification.
- Monitor the active lanes, rebalance when blockers appear, and collapse back
to serial execution when overlap emerges.
- Own the merge, final verification, and concise user summary.
- Use
talos-parallelize/scripts/parallelize_task.py for the initial lane
graph and talos-parallelize/scripts/monitor_parallel_plan.py while the
lanes are active.
- When the work will span multiple iterations, prefer
talos-parallelize/scripts/orchestrate_parallel_run.py so the lane plan and
lane states live in one run directory.
- Use the run helper's
handoffs command to generate the actual prompts for
the lane agents that should start or resume.
Guardrails:
- Do not create concurrent edits against the same file or generated output.
- Do not let one lane outrun a contract, migration, or runtime prerequisite.
- Do not claim progress without lane-level verification and one final merged
verification pass.