Redshift switch profile

Redshift Switch Profile

Flip the active Redshift profile pointer file at ~/.config/redshift-comment-mcp/active-profile. No connection field re-entry: this skill assumes the target profile is already configured (host / user / dbname in config.toml + password in OS keychain).

For single-profile users (the majority) this skill is intentionally a no-op — see "When NOT to Use".

When to Use

• User has 2+ configured Redshift profiles and wants to switch which one the MCP server connects to.
• User invokes /redshift-switch-profile <name> (direct form).
• User invokes /redshift-switch-profile (interactive — lists profiles, asks which one).
• User says "switch to dev cluster" / "切換到 prod" / "クラスタを切替" with multiple clusters configured.

When NOT to Use

• Only default (or only one profile total) → friendly bow-out, suggest /redshift-setup <name> to add a second cluster.
• Zero profiles → suggest /redshift-setup to set up first cluster.
• User wants to set up a NEW profile → use /redshift-setup.
• User wants to change a profile's host / user / dbname → use /redshift-setup <name> (it offers existing values as defaults).
• User wants to change ONLY a password → use set-password CLI.

Storage Model

This skill only edits the active-profile pointer file:

Pointer file state	Effect on MCP server startup
Absent	Server falls back to "default" (canonical single-profile state)
Present, content = <name>	Server connects with <name>
Present, content = "default"	Unusual; this skill normalizes by removing the file instead of writing "default"

Connection fields (host / user / dbname / password) and [profile.<name>] blocks in config.toml are never touched by this skill.

Flow

Step 1 — Locate plugin install path

PLUGIN_ROOT=$(ls -d "$HOME/.claude/plugins/cache/redshift-comment-mcp/redshift-comment-mcp"/*/ 2>/dev/null | sort -V | tail -1)
test -n "$PLUGIN_ROOT" || { echo "Plugin not installed." >&2; exit 1; }
echo "$PLUGIN_ROOT"

Step 2 — Inventory existing profiles

uv run --project "$PLUGIN_ROOT" redshift-comment-mcp list-profiles

The command prints one profile name per line. Branch on the count:

Case 0 — no profiles exist

Print (translate to the user's chat language; keep slash command verbatim):

你還沒設定任何 Redshift profile。
先跑 /redshift-comment-mcp:redshift-setup 建立第一個。

Stop. Do NOT touch any file.

Case 1 — exactly one profile exists

Single-profile state. Switching makes no sense. Print:

你目前只有 '<NAME>' 一個 profile，沒有切換的對象。
要新增第二個 cluster 跑 /redshift-comment-mcp:redshift-setup <名字>。

Stop.

Case 2+ — multiple profiles exist

Continue to Step 3.

Step 3 — Resolve target profile name

Path A — invoked with arg (e.g., /redshift-switch-profile prod)

Use the arg as <NAME>. Validate two things:

1. Format: matches ^[A-Za-z0-9_-]+$. If not, reject:

   '<arg>' 不是合法的 profile 名稱（只能用英數字 / 底線 / 連字號）。
   

   Stop.

2. Existence: must appear in the list-profiles output. If not:

   找不到 '<NAME>' profile。目前有：<LIST>
   要切到哪一個？(輸入名字)
   

   Wait for user reply, validate against list, retry once. On second failure, stop and ask the user to re-invoke.

Path B — invoked with no arg

Read the current active profile to inform the user:

uv run --project "$PLUGIN_ROOT" python <<'PYEOF'
from redshift_comment_mcp import config as cfg
print(cfg.read_active_profile() or "default")
PYEOF

Save the output as <CURRENT>. Then ask:

目前 active 是 '<CURRENT>'，可選的 profile：
  - <profile-1>
  - <profile-2>
  - <profile-3>
要切到哪一個？(輸入名字)

Validate the reply against the list, retry once on mismatch. On second failure, stop.

Step 4 — Idempotent check

If <NAME> == <CURRENT>, print:

'<NAME>' 已經是 active profile，沒有要切換的。

Stop. Do NOT touch any file.

Step 5 — Apply pointer file change

uv run --project "$PLUGIN_ROOT" python <<'PYEOF'
from redshift_comment_mcp import config as cfg
NAME = "<NAME>"
if NAME == "default":
    cfg.clear_active_profile()
else:
    cfg.write_active_profile(NAME)
print("✓ active profile updated")
PYEOF

Step 6 — Verify the target profile actually works

uv run --project "$PLUGIN_ROOT" \
  redshift-comment-mcp test-connection --profile "<NAME>"

Expected exit codes:

• 0 + ✓ Connected. database=X, user=Y → success. Continue to Step 7.
• 1 → connection failed. The pointer file is already updated, but the target can't connect. Tell the user the stderr and suggest:
  • check network / VPN access to the cluster
  • rotate the password via uvx --from "git+https://github.com/kouko/redshift-comment-mcp.git" redshift-comment-mcp set-password --profile <NAME>
  • or run /redshift-comment-mcp:redshift-setup <NAME> to re-do connection fields (host / user / dbname) Do NOT auto-revert the pointer file — the user may want to debug while the new profile is selected.
• 2 → profile or password missing. Means the profile isn't actually configured (defensive — list-profiles claimed it exists). Suggest /redshift-comment-mcp:redshift-setup <NAME>.

Step 7 — Print success + restart hint

✓ Active profile switched to '<NAME>'.
  下一步：重啟 Claude Code，MCP server 會用 <NAME> 連線。

(Translate prose; keep slash commands and CLI commands verbatim.)

Safety Notes

• No connection field re-entry. This skill never asks for host / user / dbname / password. To update those, use /redshift-setup <NAME> instead.
• Pointer file is mode 600, same as config.toml.
• Restart required. The MCP server reads the pointer file on startup; mid-session pointer changes have no effect until Claude Code restarts.

Reference

• Repo: https://github.com/kouko/redshift-comment-mcp
• Underlying CLI: src/redshift_comment_mcp/setup_cli.py
• Sister skill: /redshift-comment-mcp:redshift-setup (creates / edits profiles; re-collects connection fields)