---
description: Configure and manage categorized Stop-hook reflection questions
argument-hint: "list | create <category|question> ... | update <category|question> ... | delete <category|question> ..."
allowed-tools: Bash(python:*)
---

Manage the canonical reflection question bank in `.prd_plugin/config.json`.
Use the helper rather than calculating `RFQ-*` IDs or editing arrays manually.

The argument is `$ARGUMENTS`. Run the downstream helper when installed, else
the hub helper:

- downstream: `python .prd_plugin/scripts/prd_reflections.py $ARGUMENTS`
- hub: `python scripts/prd_reflections.py $ARGUMENTS`

Common operations:

- `list --json`
- `create category recovery --name "Corrective recovery"`
- `create question --category recovery --text "Would a better retry help?"`
- `update category recovery --enabled false`
- `update question RFQ-006 --enabled false`
- `update question RFQ-006 --category harness_friction`
- `delete question RFQ-006`
- `delete category recovery` (requires `--cascade` when non-empty)

Global Stop behavior remains in the validated toggle registry:

- `python .prd_plugin/scripts/prd_config.py set reflection.enabled on`
- `python .prd_plugin/scripts/prd_config.py set reflection.on_stop off`
- `python .prd_plugin/scripts/prd_config.py set reflection.max_questions_per_stop 3`

The MCP equivalents are `prd_reflection_list`, `prd_reflection_create`,
`prd_reflection_update`, and `prd_reflection_delete`. Reflection answers must be
concise observable conclusions, never private chain-of-thought or secrets.