pstack 'NSFW' prompt: make repeated repo mistakes impossible, not documented
AccBalanced · x · 2026-11-22
Ray Fernando shares a pstack prompt (/poteto-mode) for repo hygiene: have the agent read git logs, past fixes and review comments to find recurring human/agent mistakes, then make each one impossible via architecture first, then types, lint and tests. It also mandates deleting comments that restate code, encoding constraints behind important ones, replacing implementation-pinning unit tests with formal models that replay past bugs in CI, while keeping security/legal tests, shipping one verifiable PR per step.
More from coding & agent
- Dev ships ComfyUI node that runs prompts sequentially from a spreadsheet with wildcard support — ForesterAI · 2026-11-22
- Addy Osmani's agent-skills: 26 skills encoding senior-engineer workflows, 120k stars — victor_explore · 2026-11-22
- Deep dive: how cgo works at build time and its pointer-check quirks — abhi9u · 2026-11-22
- Timemator 3.4 adds MCP server with 21 tools for Claude Code, Codex and Cursor — dean_rie · 2026-11-22
- Anthropic's Claude Academy offers free official courses with completion badges — rubenhassid · 2026-11-22
- Bypassing iOS Safari's 60 FPS cap: compositor-driven scrolling unlocks 120Hz web interactions — Vjeux · 2026-11-22