feat(reflexion): implement issue-driven cloud runner

[bronz3beard]

This PR implements WS-7: Issue-driven cloud runner per the loop-engineering playbook. The core requirement is that the Reflexion Loop can be interacted with exclusively through GitHub issues without the Action ever pushing code to the repository.

Changes

1. GitHub Workflow (.github/workflows/reflexion-issue-runner.yml):

   • Triggers on issues (labeled reflexion:run), issue_comment (starting with /reflexion), and workflow_dispatch.
   • Permissions are restricted to issues: write, actions: read, contents: read.
   • Idempotency enforced via concurrency grouping by issue number.

2. Runner Script (scripts/reflexion-issue-runner.js):

   • Implements both the START path (running the loop initially and uploading state artifacts) and RESUME path (downloading the previous artifact, accepting YAML input, and resuming the loop).
   • Utilizes GitHub API (via fetch) internally since the gh CLI does not run on Node natively.
   • Extracts zips safely via adm-zip.
   • Parses incoming YAML securely using js-yaml and strict zod schema definitions.
   • Strict idempotency handled by looking back at processed-comment-id invisible HTML markers in bot responses.
   • 10-turn cap hard limits.

3. Pure Logic Scripts (scripts/reflexion-issue-runner-utils.js):

   • Handles text parsing, extraction of run markers, comment format generation, and Zod answer validation in a pure, testable environment.

4. Docs (docs/reflexion-issue-runner.md):

   • Covers secrets layout, phone workflow behavior, state isolation security models.

5. Dependencies:

   • Adds adm-zip and js-yaml.

Evidence

> @bronz3beard/tech-lead-stack@1.0.0 check-types
> tsc --noEmit --pretty --skipLibCheck

> @bronz3beard/tech-lead-stack@1.0.0 test
> jest

 PASS  scripts/__tests__/reflexion-issue-runner-utils.test.js
  reflexion-issue-runner-utils
    extractRunIdMarker
      ✓ should extract a valid marker
    extractProcessedCommentIdMarker
      ✓ should extract a valid marker
    formatRunnerComment
      ✓ should format a diagnostic error message
      ✓ should include processed-comment-id marker
    extractYamlBlock
      ✓ should extract a standard yaml block
    validateAnswers
      ✓ validates correct answers structure
      ✓ validates directive only
      ✓ fails on missing runId

---

PR created automatically by Jules for task 4741346364888340434 started by @bronz3beard

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
ai-tech-lead-stack	Ready	Preview, Comment	Jul 14, 2026 2:33pm

[google-labs-jules]

👋 Jules, reporting for duty! I'm here to lend a hand with this pull request.

When you start a review, I'll add a 👀 emoji to each comment to let you know I've read it. I'll focus on feedback directed at me and will do my best to stay out of conversations between you and other bots or reviewers to keep the noise down.

I'll push a commit with your requested changes shortly after. Please note there might be a delay between these steps, but rest assured I'm on the job!

For more direct control, you can switch me to Reactive Mode. When this mode is on, I will only act on comments where you specifically mention me with @jules. You can find this option in the Pull Request section of your global Jules UI settings. You can always switch back!

New to Jules? Learn more at jules.google/docs.

---

For security, I will only act on instructions from the user who triggered this task.