---
name: source-command-prd-off
description: Switch PRD Plugin completely off in this repo (hooks, stop guard, gates, workflows, telemetry) — one reversible profile. Host-native /prd-off equivalent for hosts that load skills instead of Claude command files; trigger on "/prd-off", "prd-off", or a request to run the prd-off operation.
---

# /prd-off (host-native command skill)

Generated from `commands/prd-off.md` by `scripts/generate_command_skills.py` — do not edit by hand; edit the command file and regenerate.

Treat the user's invocation as this command. Where the instructions use `$ARGUMENTS`, substitute the arguments the user supplied.

Apply the `off` configuration profile, then report plainly.

Run (downstream path first, hub fallback):

- downstream: `python .prd_plugin/scripts/prd_config.py profile apply off`
- hub: `python scripts/prd_config.py profile apply off`

Then tell the user, in plain language:

1. Every automatic PRD Plugin behavior in this repo is now off: no session
   nudges, no run-until-done stop guard, no commit gate, no drift/reflection/
   telemetry, no deterministic lifecycle workflows.
2. Skills, commands, and MCP tools stay installed but nothing fires on its own;
   the method only acts if explicitly invoked.
3. Reverse it any time with `/prd-on` (balanced profile) or
   `prd_config.py profile apply <lean|balanced|full>`.
4. Honest limits: instructions already loaded into the CURRENT session stay in
   context until a new session starts, and CLAUDE.md/AGENTS.md files remain in
   the repo. For full host-level removal (skills disappear from the app too),
   the user runs: `claude plugin disable prd-plugin@prd-plugin` (Claude) or
   toggles the plugin off in Codex Desktop — both take effect on app restart.

Do nothing else.

## Staleness

This adapter owns no records of its own; apply the shared policy in `.prd_plugin/method/staleness-rules.md` to any records the underlying operation reads or writes.