{"playbook": "coding-agents", "title": "Working with AI coding agents", "description": "Getting reliable work out of coding agents: instruction and context files, task scoping, reviewing agent-written code, repository conventions, hooks and workflow guardrails.", "url": "https://wpnews.pro/playbooks/coding-agents", "count": 5, "recommendations": [{"advice": "Add a `verify` skill in `.claude/skills/` that starts your app or simulator, request property tests for round-trip and idempotence, and stop for a human after three failed rounds.", "takeaway": "A separate grader checking agent output against a rubric raised task success by up to 10 points in an internal test, and verification loops let the Claude Code team run dozens or hundreds of Claudes in parallel instead of one or two.", "article_url": "https://wpnews.pro/news/make-your-agent-prove-its-work", "title": "Make your agent prove its work", "source_url": "https://www.bulaev.net/p/make-your-agent-prove-its-work", "published_at": "2026-11-17T19:43:52+00:00"}, {"advice": "Write the naming and folder convention into the agent's prompt as an explicit rule, including the reason behind it, then have the coding agent apply it across the repo in one pass.", "takeaway": "Delegating a 750-file, 2,148-line rename and reorganization to Claude Code paid off because the task was too low-value to justify days of manual work; encoding the rule that folder location specifies file role (plans/ipad.md, prompts/ipad.md) let the agent apply it consistently.", "article_url": "https://wpnews.pro/news/slop-is-a-choice", "title": "Slop Is a Choice", "source_url": "https://racecondition.software/blog/slop-is-a-choice/", "published_at": "2026-11-17T08:00:00+00:00"}, {"advice": "Start with a PR factory reviewing each agent pull request individually, then batch dependency-mapped tickets into a release factory with one combined staging review before deploy.", "takeaway": "Batch related, dependency-mapped tickets so parallel coding agents build in one release and a human reviews the combined output in staging, since a meter.org study found roughly half of test-passing agent pull requests were rejected by maintainers.", "article_url": "https://wpnews.pro/news/how-to-build-a-release-factory-for-ai-coding-agents", "title": "How to Build a Release Factory for AI Coding Agents", "source_url": "https://www.mindstudio.ai/blog/how-to-run-ai-software-factory-release/", "published_at": "2026-11-17T00:00:00+00:00"}, {"advice": "Have your coding agent read your GitHub Actions run history, list CI optimizations in a table ranked by estimated minutes saved per day, then ship the top fixes as reviewable PRs.", "takeaway": "Ranking CI fixes by estimated minutes saved per day from actual GitHub Actions run history, as OpenAI's team did with Codex, produced about 30 PRs that cut median CI time 55% from roughly 10 minutes to 4.5 minutes.", "article_url": "https://wpnews.pro/news/openai-s-ci-playbook-for-github-actions", "title": "OpenAI's CI playbook (for GitHub Actions)", "source_url": "https://starsling.dev/blog/openai-ci", "published_at": "2026-10-22T00:00:00+00:00"}, {"advice": "Scope CI to the tests each change can affect and hand coding agents that specific test set so they can self-verify; Anthropic used this to absorb a 25x rise in CI jobs.", "takeaway": "Run only the tests a change can affect via test impact analysis instead of every test on every PR, and give coding agents a specific set of valid tests so they can self-verify and iterate — Anthropic's approach held up as CI jobs grew 25x in six months.", "article_url": "https://wpnews.pro/news/anthropic-s-ci-playbook-as-github-actions-prompts-for-your-agent", "title": "Anthropic's CI playbook, as GitHub Actions prompts for your agent", "source_url": "https://starsling.dev/blog/anthropic-ci", "published_at": "2026-10-19T00:00:00+00:00"}]}