Clarify stuck eval completion criteria

evals/prompts/stuck-detection.txt

@@ -32,6 +32,7 @@ First, check Tool Calls for any "running" or "pending" status. If found, return:
 - Agent shows test results with "passed" or "✓"  
 - Agent's response indicates task fulfilled: "I've added", "I've fixed", "The X now works"
 - No pending work mentioned (no "Next I will...", "Still need to...")
+- Use ONLY when there is evidence of completion. If the response is purely a plan or question, do not return "complete".
 - IMPORTANT: If the user's task requires code changes (fix, implement, add, create, build, refactor), message_completed is true, and the Tool Calls show ONLY read operations (read, glob, grep, git log/status/diff, webfetch, task/explore) with NO write operations (edit, write, bash with build/test/commit, PR creation), the task is NOT complete — classify as genuinely_stuck with stuck=true. Analyzing or recommending changes is not the same as implementing them.
 - If this rule applies, do NOT return reason "complete".
 → reason: "complete"

package.json

@@ -15,12 +15,13 @@
     "test:e2e": "node --import tsx --test test/e2e.test.ts",
     "test:tts:manual": "node --experimental-strip-types test/tts-manual.ts",
     "test:load": "node --import tsx --test test/plugin-load.test.ts",
-    "test:reflection-3": "node --import tsx --test test/reflection-static.eval.test.ts",
+    "test:reflection": "node --import tsx --test test/reflection-static.eval.test.ts",
     "test:install": "jest test/install.test.ts --testTimeout=120000",
     "typecheck": "npx tsc --noEmit",
+    "install": "npm run install:telegram && npm run install:tts && npm run install:reflection && npm run install:sentry && npm run install:opencode-worktree",
     "install:telegram": "mkdir -p ~/.config/opencode/plugin && cp telegram.ts sentry.ts ~/.config/opencode/plugin/ && node scripts/ensure-deps.js && cd ~/.config/opencode && bun install",
     "install:tts": "mkdir -p ~/.config/opencode/plugin && cp tts.ts sentry.ts ~/.config/opencode/plugin/ && node scripts/ensure-deps.js && (cd ~/.config/opencode && bun install) && bash scripts/setup-coqui.sh",
-    "install:reflection-3": "mkdir -p ~/.config/opencode/plugin && cp reflection-3.ts ~/.config/opencode/plugin/reflection.ts && cp sentry.ts ~/.config/opencode/plugin/ && node scripts/ensure-deps.js && cd ~/.config/opencode && bun install",
+    "install:reflection": "mkdir -p ~/.config/opencode/plugin && cp reflection-3.ts ~/.config/opencode/plugin/reflection.ts && cp sentry.ts ~/.config/opencode/plugin/ && node scripts/ensure-deps.js && cd ~/.config/opencode && bun install",
     "install:sentry": "mkdir -p ~/.config/opencode/plugin && cp sentry.ts ~/.config/opencode/plugin/ && node scripts/ensure-deps.js && cd ~/.config/opencode && bun install",
     "install:opencode-worktree": "mkdir -p ~/.local/bin && cp scripts/opencode-worktree.ts ~/.local/bin/opencode-worktree && chmod +x ~/.local/bin/opencode-worktree",
     "eval": "cd evals && npx promptfoo eval",

test/telegram.test.ts

@@ -586,7 +586,7 @@ describe("Error Handling", () => {
     const result = await response.json()
     // Either text_sent is false OR error is present
     expect(result.text_sent === false || result.error).toBeTruthy()
-  }, 10000) // Extended timeout for network latency
+  }, 20000) // Extended timeout for network latency
 
   it("webhook handles malformed JSON gracefully", async () => {
     const response = await fetch(WEBHOOK_URL, {