Clickui Code

ClickUI Code Review

Before running risky shell commands, get user approval via this skill. Always show the user exactly which files will change and what the diffs look like — this is the default, not optional.

Preferences

Before submitting a command for review, fetch user preferences and apply any rules under ## Code Review:

curl -s "$AGENTCLICK_BASE/api/preferences/style"

Step 1: Generate the diff

Before submitting, capture the exact changes so the user can see them in the review UI.

# Changes a patch would make (before applying):
git diff --unified=3 HEAD -- path/to/file.ts

# All staged changes:
git diff --cached --unified=3

# What a specific commit changed:
git show <commit> --unified=3

# Multiple files at once:
git diff --unified=3 HEAD -- src/api/client.ts src/utils/retry.ts

Escape each diff as a JSON string: replace newlines with \n and double-quotes with \".

Step 2: Submit the command for review

if curl -s --max-time 1 http://localhost:38173/api/health > /dev/null 2>&1; then
  AGENTCLICK_BASE="http://localhost:38173"
else
  AGENTCLICK_BASE="http://host.docker.internal:38173"
fi

RESPONSE=$(curl -s -X POST "$AGENTCLICK_BASE/api/review" \
  -H "Content-Type: application/json" \
  -d '{
    "type": "code_review",
    "sessionKey": "'"$SESSION_KEY"'",
    "payload": {
      "command": "THE_EXACT_COMMAND_YOU_WANT_TO_RUN",
      "cwd": "WORKING_DIRECTORY",
      "explanation": "WHAT_THIS_COMMAND_DOES_AND_WHY",
      "risk": "low|medium|high",
      "affectedFiles": [
        {
          "path": "backend/src/utils/retry.ts",
          "status": "added",
          "diff": "@@ -0,0 +1,12 @@\n+export async function retry<T>(fn: () => Promise<T>, times = 3): Promise<T> {\n+  let last: unknown\n+  for (let i = 0; i < times; i++) {\n+    try { return await fn() } catch (e) { last = e }\n+  }\n+  throw last\n+}"
        },
        {
          "path": "backend/src/api/client.ts",
          "status": "modified",
          "diff": "@@ -1,8 +1,9 @@\n import axios from '''axios'''\n+import { retry } from '''../utils/retry'''\n \n export async function fetchUser(id: string) {\n-  return axios.get(`/users/${id}`)\n+  return retry(() => axios.get(`/users/${id}`))\n }"
        },
        {
          "path": "backend/src/api/legacyClient.ts",
          "status": "deleted",
          "diff": "@@ -1,5 +0,0 @@\n-// deprecated — use client.ts\n-import axios from '''axios'''\n-export const get = (url: string) => axios.get(url)"
        },
        {
          "path": "backend/src/api/index.ts",
          "status": "renamed",
          "oldPath": "backend/src/api/exports.ts",
          "diff": "@@ -1,2 +1,2 @@\n-export * from '''./legacyClient'''\n+export * from '''./client'''"
        }
      ]
    }
  }')
SESSION_ID=$(echo "$RESPONSE" | grep -o '"sessionId":"[^"]*"' | cut -d'"' -f4)
echo "Session: $SESSION_ID"

affectedFiles entry fields

If you cannot generate a diff (e.g. the command doesn't touch tracked files), omit diff and the file will still appear in the mind-map tree without a diff panel.

Legacy fallback: If you only have a flat list of paths and no diffs, use "files": ["backend/src/index.ts", ...] instead of affectedFiles. Paths must still be relative to project root. The UI will show the tree without diff panels.

Step 3: Poll for decision

# Detect environment: GOG_ACCOUNT is set in Docker (docker-compose.yml), absent elsewhere
if [ -n "$GOG_ACCOUNT" ]; then
  # Docker: use short-poll (one curl per exec, you are the loop controller)
  curl -s "$AGENTCLICK_BASE/api/sessions/${SESSION_ID}"
else
  # Non-Docker: use blocking /wait
  curl -s --max-time 310 "$AGENTCLICK_BASE/api/sessions/${SESSION_ID}/wait"
fi

After each poll:

• If status is "completed" → proceed to Step 4.
• If status is still "pending" → wait 1 second (sleep 1 as a separate exec in Docker), then poll again.

In Docker (GOG_ACCOUNT set): run ONE curl per exec call. Do NOT use a bash while loop. You are the loop controller. In non-Docker: the /wait call blocks until the user approves or rejects, then process the result.

Step 4: Act on the decision

• result.approved: true → Run the command immediately. Do NOT ask the user again. The user already approved in the UI. If result.note is set, adjust the command accordingly.
• result.approved: false → Do not run the command. Inform the user.