Auto Dev

Pick an open issue, develop a fix, and submit a PR.

When triggered

This skill runs as a daily cron task. Process one issue per run.

Step 1: Find an issue to work on

bash

gh issue list --repo l1veIn/nanobot-auto --label auto-report --state open --json number,title,body --limit 1

If no issues found, stop — nothing to do.

Step 2: Analyze the issue

Read the issue title and body. Understand:

•What is the problem?
•Which file(s) need to be changed?
•What is the expected fix?

Step 3: Prepare workspace

bash

git checkout main
git pull origin main
git checkout -b fix/issue-<NUMBER>

Step 4: Develop the fix using Codex

Use OpenAI Codex in YOLO mode (full-auto, no confirmation prompts) to implement the fix:

bash

codex --approval-mode full-auto "Read issue #<NUMBER> from this repo and fix it. Run tests after making changes."

Codex will autonomously read the issue, modify files, and run verification.

If Codex is not available, fall back to manual development:

•Use read_file to examine relevant source files
•Use write_file or edit_file to make changes
•Minimal changes only — fix the reported issue, nothing else

Step 5: Verify

Run syntax check on every modified .py file:

bash

python -m py_compile <modified_file.py>

If any file fails compilation, revert the change and try a different approach.

If tests exist, run them:

bash

python -m pytest tests/ -x --tb=short 2>&1 | tail -30

If tests fail, investigate and fix. If you cannot fix tests, abandon this issue.

Step 6: Commit and push

bash

git add -A
git commit -m "fix: <concise description> (closes #<NUMBER>)"
git push origin fix/issue-<NUMBER>

Step 7: Create PR

bash

gh pr create --repo l1veIn/nanobot-auto \
  --title "fix: <concise description>" \
  --body "Closes #<NUMBER>

## Changes
- <list what was changed and why>

## Verification
- py_compile: ✅
- pytest: ✅ / ⚠️ no tests / ❌ (details)"

Step 8: Clean up

bash

git checkout main