[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$fjm5incwip8pa":3},{"code":4,"message":5,"data":6,"timestamp":46},10000,"success",{"full_name":7,"position":8,"description":9,"language":10,"stars":11,"forks":12,"stars_gained":13,"age_days":14,"hn_threads":15,"one_liner":16,"summary_md":17,"topic":18,"github_topics":21,"license":42,"homepage":43,"updated_at":44,"history":45},"michael-denyer\u002Fpstack-claude",7,"Claude Code, Codex, Copilot, Pi, OpenCode, Muse, Gemini, and Prime Agent versions of Poteto's pstack. Rigorous agent workflows with Cursor primitives translated for other harnesses.","JavaScript",3370,314,2792,151,[],"A port of Cursor's opinionated skill stack that turns any coding agent into a verified, workflow-driven worker.","## The problem\n\nCoding agents are fast but unfocused: given a vague request they guess at scope, skip reproduction, and declare work done without evidence. Lauren Tan's pstack addressed this inside Cursor by encoding an opinionated workflow into skills. This repository ports that stack to Claude Code, Codex, GitHub Copilot, Pi, Muse, OpenCode, Gemini CLI and Prime Agent, so teams on other harnesses get the same routing and verification discipline.\n\n## What's different\n\nThe usual approach is a pile of prompt files and a convention that developers remember to invoke the right one. pstack's core is `poteto-mode`, a router: you state a goal, and it selects a playbook (bug fix, refactor, performance investigation, PR maintenance, shipping) and drives sub-skills like `how`, `why`, `architect`, `arena`, `swarm` and `interrogate` on your behalf. Two things stand out. First, evidence is built into the loop: a bug fix must rerun the failing case and hand back failing and passing output. Second, an optional formal-verification skill targets concurrency bugs and invariants that tests cannot reach, generating a reusable project skill for the next agent. Installation mechanics are deliberately harness-native — a plugin marketplace on Claude Code, Codex and Copilot, a package on Pi, symlinked skills into `~\u002F.agents\u002Fskills\u002F` elsewhere.\n\n## Who it's for\n\nDevelopers already using a coding agent daily, on a codebase with tests, who want repeatable process rather than another prompt collection — especially teams standardizing agent behavior across mixed harnesses. Solo scripters, people on harness versions the port hasn't been tested against, and anyone wanting a lightweight single-prompt helper should look elsewhere; the workflow is opinionated and heavy by design.\n\n## Cost of adoption\n\nRequires a supported agent version (Claude Code 2.1.292 or later; Copilot CLI is only tested on 1.0.87–1.0.92, with a documented skill-discovery bug on 1.0.92). Hooks must be trusted or approved before routing works, and harnesses without a hook need a standing instruction to activate `poteto-mode`. Model defaults are per-role and configured through `setup-pstack`. MIT licensed, no server or telemetry, scripts run locally. This is a port that tracks upstream and carries declared forks, so staying current means re-pulling and re-linking.\n\n## Verdict\n\nThis is a brand-new project — created 2026-05-26 — yet it drew +2,792 stars in a single month, landing at #7 on the monthly board, with no Hacker News discussion behind the spike. That pattern points to a launch that found an audience quickly rather than a long-established tool, and the README itself notes version-specific Copilot breakage. Its value depends entirely on whether Lauren Tan's upstream workflow suits you; the port's own contribution is breadth of harness support. Worth trying if you already run a coding agent daily, but expect to chase upstream changes and harness quirks.",{"slug":19,"name":20},"developer-tools","Developer Tools",[22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41],"agent-plugin","agent-skills","agentic-ai","anthropic","chatgpt","claude","claude-code","claude-code-skills","claude-plugin","claude-skills","code-review","codex","codex-cli","codex-plugin","codex-skills","coding-agent","gemini-cli","muse-plugin","opencode","pstack","MIT",null,"2026-10-24T16:36:41Z",[],1792868478]