gstack/test/skill-e2e-plan-ceo-plan-mode.test.ts

/**
 * plan-ceo-review plan-mode smoke (gate, paid, real-PTY).
 *
 * Asserts: when /plan-ceo-review is invoked in plan mode, the FIRST terminal
 * outcome is 'asked' — a skill-question numbered list. Permission dialogs
 * (which also render numbered lists) are filtered out by `runPlanSkillObservation`
 * via its `isPermissionDialogVisible(visible.slice(-1500))` short-circuit.
 *
 * Reaching 'plan_ready' first IS the regression we want to catch: the agent
 * skipped Step 0 entirely and went straight to ExitPlanMode. The original
 * failure had the assistant read a diff, write a plan with two issues, and
 * call ExitPlanMode without ever firing AskUserQuestion — the user had to
 * manually call out the missing per-issue questions.
 *
 * Why this skill is special: unlike plan-eng-review / plan-design-review /
 * plan-devex-review (whose smokes accept either 'asked' or 'plan_ready'),
 * plan-ceo-review's template mandates Step 0A premise challenge (3 baked-in
 * questions) AND Step 0F mode selection BEFORE any plan write. There is no
 * legitimate path to plan_ready that does not first emit a skill-question
 * numbered prompt.
 *
 * Env passthrough: passes `QUESTION_TUNING=false` and `EXPLAIN_LEVEL=default`
 * via the runner's env option. Today these are advisory — `gstack-config`
 * reads `~/.gstack/config.yaml`, not env vars, so a contributor with
 * `question_tuning: true` set in their YAML config can still see AUTO_DECIDE
 * masking. The env passthrough is wired so a future gstack-config change to
 * honor env overrides will make this test hermetic without further edits.
 * Tracked as a post-merge follow-up.
 *
 * FAIL conditions: 'plan_ready' first, silent Write/Edit before any prompt,
 * claude crash, timeout.
 *
 * See test/helpers/claude-pty-runner.ts for runner internals.
 */

import { describe, test, expect } from 'bun:test';
import {
  runPlanSkillObservation,
  planFileHasDecisionsSection,
  assertReportAtBottomIfPlanWritten,
  isProseAUQVisible,
} from './helpers/claude-pty-runner';

const shouldRun = !!process.env.EVALS && process.env.EVALS_TIER === 'gate';
const describeE2E = shouldRun ? describe : describe.skip;

describeE2E('plan-ceo-review plan-mode smoke (gate)', () => {
  test('first terminal outcome is asked (Step 0 fires before any plan write)', async () => {
    const obs = await runPlanSkillObservation({
      skillName: 'plan-ceo-review',
      inPlanMode: true,
      timeoutMs: 300_000,
      env: { QUESTION_TUNING: 'false', EXPLAIN_LEVEL: 'default' },
    });

    if (obs.outcome !== 'asked') {
      const diagnosis =
        obs.outcome === 'plan_ready'
          ? `'plan_ready' first means the agent skipped Step 0 entirely and went straight to ExitPlanMode without asking.`
          : obs.outcome === 'timeout'
            ? `Timeout means the agent neither asked nor completed within the budget — likely hung mid-question or stuck on a permission dialog.`
            : obs.outcome === 'silent_write'
              ? `Silent Write/Edit fired to an unsanctioned path before any AskUserQuestion — also a Step 0 skip.`
              : `Outcome '${obs.outcome}' is unexpected; investigate the evidence below.`;
      throw new Error(
        `plan-ceo-review smoke FAILED: outcome=${obs.outcome}\n` +
          `${diagnosis}\n` +
          `Expected 'asked'. See plan-ceo-review/SKILL.md.tmpl: the Step 0 STOP rules ` +
          `and the "One issue = one AskUserQuestion call" rule under "CRITICAL RULE — ` +
          `How to ask questions".\n` +
          `summary: ${obs.summary}\n` +
          `elapsed: ${obs.elapsedMs}ms\n` +
          `--- evidence (last 2KB visible) ---\n${obs.evidence}`,
      );
    }
    assertReportAtBottomIfPlanWritten(obs);
  }, 360_000);

  // v1.21+ regression: Conductor launches Claude Code with
  // `--disallowedTools AskUserQuestion --permission-mode default` (verified
  // via `ps` on the live Conductor claude process). Native AskUserQuestion
  // is removed from the model's tool registry; without fallback guidance
  // the model can't ask and silently proceeds.
  //
  // After v1.28+ (forever-war fix), the preamble fallback that wrote a
  // "## Decisions to confirm" section was deleted in favor of a hard
  // BLOCKED rule. The pass envelope under --disallowedTools accepts:
  //   - 'asked'      — model emits a numbered-option prompt as prose
  //   - 'plan_ready' WITH (## Decisions section [legacy]
  //                  OR BLOCKED string visible [post-fix])
  //   - 'exited'     WITH BLOCKED string visible [post-fix]
  //
  // The legacy `## Decisions` path stays in the envelope so this test
  // keeps passing during the migration window when the fallback delete
  // and resolver edits land in the same PR but mid-rebase states are
  // possible. Once the deletion has been on main long enough that the
  // generated SKILL.md cache has flushed, the legacy branch can be
  // removed in a follow-up.
  //
  // Failure signals (regression we DO want to catch):
  //   - 'auto_decided' — AUTO_DECIDE preamble fired without /plan-tune opt-in
  //   - 'silent_write' — Write/Edit before any AUQ surface
  //   - 'timeout'      — neither asked nor terminated in budget
  //   - 'plan_ready' or 'exited' WITHOUT either Decisions section or BLOCKED
  test('AskUserQuestion surfaces when --disallowedTools AskUserQuestion is set', async () => {
    const obs = await runPlanSkillObservation({
      skillName: 'plan-ceo-review',
      inPlanMode: true,
      extraArgs: ['--disallowedTools', 'AskUserQuestion'],
      timeoutMs: 300_000,
    });

    // The user must SEE the question one way or another. Three valid surfaces:
    //   1. `## Decisions to confirm` section in the plan file (legacy fallback)
    //   2. `BLOCKED — AskUserQuestion` string visible in TTY (post-v1.28 BLOCKED rule)
    //   3. Numbered/lettered options visible in TTY as prose (post-v1.28 prose-AUQ rendering)
    const blockedVisible = /BLOCKED\s*[—-]\s*AskUserQuestion/i.test(obs.evidence);
    const proseAUQVisible = isProseAUQVisible(obs.evidence);
    const surfaceVisible = blockedVisible || proseAUQVisible;

    if (
      obs.outcome === 'auto_decided' ||
      obs.outcome === 'silent_write' ||
      obs.outcome === 'timeout'
    ) {
      throw new Error(
        `plan-ceo-review AskUserQuestion-blocked regression: outcome=${obs.outcome}\n` +
          `summary: ${obs.summary}\n` +
          `elapsed: ${obs.elapsedMs}ms\n` +
          `--- evidence (last 2KB visible) ---\n${obs.evidence}`,
      );
    }
    if (obs.outcome === 'exited' && !surfaceVisible) {
      throw new Error(
        `plan-ceo-review AskUserQuestion-blocked regression: outcome=exited without any visible question surface (no BLOCKED string, no prose-rendered AUQ options). Model quit silently.\n` +
          `--- evidence (last 2KB visible) ---\n${obs.evidence}`,
      );
    }
    if (obs.outcome === 'plan_ready') {
      if (!obs.planFile) {
        if (!surfaceVisible) {
          throw new Error(
            `plan-ceo-review AskUserQuestion-blocked regression: outcome=plan_ready but no plan file path detected, no BLOCKED string, no prose AUQ options. Cannot verify the model used any legitimate path.\n` +
              `--- evidence (last 2KB visible) ---\n${obs.evidence}`,
          );
        }
      } else if (!planFileHasDecisionsSection(obs.planFile) && !surfaceVisible) {
        throw new Error(
          `plan-ceo-review AskUserQuestion-blocked regression: model wrote ${obs.planFile} without a "## Decisions" section AND no BLOCKED string AND no prose AUQ options in TTY. Step 0 was silently skipped.\n` +
            `--- evidence (last 2KB visible) ---\n${obs.evidence}`,
        );
      }
    }
    expect(['asked', 'plan_ready', 'exited']).toContain(obs.outcome);
    // NOTE: assertReportAtBottomIfPlanWritten is intentionally NOT called
    // here. This test runs --disallowedTools AskUserQuestion and only
    // checks "did the question surface" — the model can't run the full
    // multi-section review without AUQ tools, so no review report exists
    // to enforce the at-bottom contract against. The contract is
    // exercised by the periodic finding-count tests, which DO run the
    // full review.
  }, 360_000);
});