/setup

name: subscope-setup
description: Reconfigure individual pieces of subscope after first-run onboarding. Use this when you want to add or change LLM provider, swap a destination (Notion / Slack / Obsidian), or rotate credentials. NOT the first-run flow. First-time users should run /subscope-onboard instead. Triggers on "/subscope-setup", "reconfigure subscope", "change subscope settings", "update subscope LLM", "rotate subscope credentials", "swap subscope destination".
allowed-tools: Bash, Read, Write, Edit

/subscope-setup

Reconfiguration tool for an already-onboarded subscope install. For first-run, route the user to `/subscope-onboard` instead. The 6 steps below are now individually addressable: ask the user which one they want to change before walking the whole flow.

Goal

Let an existing user change one piece of their config without re-doing the whole onboarding.

Procedure

First, check whether the user has already onboarded:

[ -f ~/.config/subscope/subreddits.yml ] && echo "ONBOARDED" || echo "NOT_ONBOARDED"

If `NOT_ONBOARDED`: redirect to `/subscope-onboard` and stop. Do not run the steps below for a new install.

If `ONBOARDED`: ask which piece they want to change:

> What do you want to reconfigure? > 1. LLM provider (add or swap) > 2. Targeting (re-pick preset, or re-run /subscope-onboard for full re-do) > 3. Notion database > 4. Slack webhook > 5. Obsidian vault > > Pick a number, or type "all" to walk every step.

Then jump to that step only. Each step is independent.

Step 1 — LLM provider (optional)

Tell the user:

> subscope has three classification tiers: > 1. **Default:** regex-only gate, zero cost, no API key needed. > 2. **Interactive:** `/subscope-judge <surface-id>` uses your Claude Code subscription, free. > 3. **Bulk LLM:** any OpenAI-compatible provider (Anthropic, OpenAI, Groq, OpenRouter, Together, local Ollama). ~$0.50/day at 5K posts. > > Want to enable the bulk-LLM tier? (yes / skip)

If yes, the user sets one env var (`LLM_API_KEY`, or legacy `ANTHROPIC_API_KEY` / `OPENAI_API_KEY`). The base URL is auto-detected from the key prefix (`sk-ant-` → Anthropic /openai/v1, `sk-or-` → OpenRouter, `gsk_` → Groq, else OpenAI). For local Ollama or a custom endpoint, also set `LLM_BASE_URL`.

To lock the preference, write a minimal `~/.config/subscope/llm.json`:

{"provider": "openai_compatible"}

Test the resolved config (this will also fail loudly if `LLM_BASE_URL` targets a private IP — the SSRF guard is intentional):

cd "$CLAUDE_PLUGIN_ROOT" && PYTHONPATH=engine python3 -c "
from subscope.lib import classify
import json
print(json.dumps(classify.status(), indent=2))
"

If skip: write `{"provider": "disabled"}` to make the choice explicit.

Step 2 — Targeting (route to /onboard)

Don't pick a preset directly here. Tell the user:

> Targeting works best with 3 quick questions — /subscope-onboard takes about 60 seconds and produces a config tuned to your specific work, not a generic lane. Recommended for almost everyone. > > If you really want the 30-second generic lane: type `preset` and I'll show the 4 options.

If user opts for `/subscope-onboard`: invoke it. The onboard skill handles preset escape internally (`/subscope-onboard preset`) for users who type `preset` mid-flow.

If user explicitly wants preset here in setup (rare), show the 4 options + copy chosen preset to `~/.config/subscope/`. But default is route to onboard.

Research backing (Phase 9.5 validation): generic preset alone produces 3/10 ICP-match per surface. The 3-question routing pushes it to 5-7/10. Don't deprive the user of that lift unless they explicitly opt out.

Step 3 — Where do you want to see your daily surfaces?

Ask the user (this question is non-skippable, but every choice is valid):

> How do you want to see your daily surfaces? > a) **Inline table in Claude Code chat** (default, no setup, click links from chat) > b) **Notion database** (5-min setup, persistent triage across devices, queryable views) > c) **Both** (table rendered in chat AND synced to Notion) > d) **Skip** (engine prints JSON only — for piping to your own tools)

Translate the choice to the surface.yml config:

# Pick ONE based on user's answer
MODES="[table]"            # (a) default
MODES="[notion]"           # (b)
MODES="[table, notion]"    # (c)
MODES="[]"                 # (d)

cd "$CLAUDE_PLUGIN_ROOT" && cat <<EOF | PYTHONPATH=engine python3 -m scripts.write_surface_config
modes: $MODES
default_render: table
EOF

If the user picked (b) or (c), continue to Step 3b (Notion setup) below. Otherwise skip Step 3b and continue to Step 4 (Obsidian).

Step 3b — Notion DB hookup (only if Step 3 picked Notion)

If yes: 1. Ask if they have a Notion API key + DB ID, OR want to use an existing DB URL. 2. If they paste a Notion DB URL, extract the 32-char ID from it. 3. **Write the config file FIRST** (never pass `NOTION_API_KEY=$KEY` inline on the command line — the secret would land in /proc/<pid>/environ, ps output, and shell history). Use the atomic helper:

cd "$CLAUDE_PLUGIN_ROOT" && cat <<EOF | PYTHONPATH=engine python3 -m scripts.write_notion_config
api_key: $NOTION_API_KEY
database_id: $NOTION_DB_ID
EOF

The helper writes to `~/.config/subscope/notion.yml` with `chmod 600` from the moment the file appears (atomic, no umask race).

4. Test with a dry-run migration that READS the file (no env-var leak):

cd "$CLAUDE_PLUGIN_ROOT" && PYTHONPATH=engine python3 engine/scripts/notion_admin.py migrate --dry-run

5. If dry-run succeeds, run the live migration to add Pattern/State/Fit properties + backfill existing rows:

cd "$CLAUDE_PLUGIN_ROOT" && PYTHONPATH=engine python3 engine/scripts/notion_admin.py migrate

Step 4 — Obsidian (optional)

Ask:

> Want weekly pulse digests written to your Obsidian vault? (yes / skip)

If yes: 1. Ask for the absolute vault path. 2. Ask for the pulse sub