# Research Questions: Robust Pi subagent workflow for Nimrod

## Phase Contract

Allowed reads:
- `tickets/active/2026-06-03-robust-pi-subagents.md`
- `tickets/artifacts/2026-06-03-robust-pi-subagents/00-context.md`
- Pi extension/SDK docs and relevant examples
- Existing Nimrod QRSPI docs

Will write:
- `tickets/artifacts/2026-06-03-robust-pi-subagents/01-questions.md`
- next phase: `02-research.md`

## Context for Research

Investigate how Pi can run isolated helper agents for project workflow tasks, using project-local configuration where practical. Focus on extension mechanics, subprocess/SDK options, prompt/tool isolation, artifact handoff, and safety controls.

## Questions

1. How does Pi's example `subagent` extension create isolated agent contexts, configure tools/models/prompts, stream results, and return compressed output?
2. What extension APIs are needed for a Nimrod-specific subagent workflow: custom tools, commands, prompt injection, active tool control, session management, and UI confirmation?
3. What security/safety boundaries does Pi's subagent example use for user-level vs project-local agents, and what should Nimrod require?
4. How should QRSPI artifacts integrate with subagents so each subagent reads narrow inputs and writes bounded markdown outputs?
5. Should the MVP copy/adapt Pi's example extension directly, wrap it, or build a simpler project-local extension from scratch?
6. Which agent roles are needed first for Nimrod: locator, researcher, analyzer, pattern-finder, designer, planner, implementer, reviewer?
7. What verification steps can prove the subagent system actually reduces main-context load and remains safe?

## Approval Notes

- Initial questions drafted from the user's concern about context growth and the discovered Pi subagent example.