---
title: "tstack"
description: "Taimoor's drift-prevention stack: plan-guard gates plans before work, drift-audit measures running projects against their original goal, plus personal workstation skills."
canonical: https://agentpluginsdirectory.com/plugins/tstack--deepextrema-2
last-updated: 2026-10-21
---

# tstack
Taimoor's drift-prevention stack: plan-guard gates plans before work, drift-audit measures running projects against their original goal, plus personal workstation skills.
- Slug: tstack--deepextrema-2
- Publisher: Taimoor (DeepExtrema)
- Repository: https://github.com/DeepExtrema/tstack
- Manifest: plugin.json
- Version: 0.1.0
- License: MIT
- Category (editorial): other
- Skills: 8 (drift-audit, feynman-headless-research, nimblemind-greenfield-orientation, omarchy-end4-hybrid-bar, omarchy-end4-sidebar-dashboard, plan-guard, quickshell-pointer-following-dock, tesseract-omp-integration)
- MCP servers: 0
- Stars: 0
- Repository created: 2026-10-19
- Repository last pushed: 2026-10-19
- Publisher type: User
- Listing: https://agentpluginsdirectory.com/plugins/tstack--deepextrema-2
- Schema: https://agent-plugins.org/schemas/1.0.0/plugin.schema.json

## What tstack does, in the publisher's words

Skills that keep coding agents on the plan. Built after a $10 form-filling bot turned into a six-week, 18k-line root installer with six Linux users, signed documents, 46 sudo prompts, 274 continuation prompts, and zero real submissions.

Two core skills:

Both are grounded in skills/plan-guard/references/rules.md: 70 rules, each with its source and a verbatim quote or number, drawn from Anthropic's and OpenAI's prompting guides, four research papers (Overeager Coding Agents, Asymmetric Goal Drift, Evaluating Goal Drift, IFScale), Lauren Tan's pstack principles, Matt Pocock's skills, Superpowers, gstack, HumanLayer, and the talks by Dex Horthy, Boris Cherny, Simon Willison, Jesse Vincent, Geoffrey Huntley and others. Retractions and disagreements between sources are recorded there too.

From the project README, punctuation lightly normalized. Full text: https://raw.githubusercontent.com/DeepExtrema/tstack/HEAD/README.md

## Skills

- drift-audit: Audit a project that is already underway for goal drift and over-engineering; use when asked whether a project has derailed, is over-engineered, has lost its original goal, or needs its plan fixed, and whenever a project has run more than twice its estimate without the deliverable. Produces an evid…
- feynman-headless-research: Launch the real Feynman research CLI from OMP, enforce Sol 6.1-only research models, handle native plan approval, and verify report and actual child-model provenance.
- nimblemind-greenfield-orientation: Locate authoritative Nimblemind greenfield workflow code and acceptance evidence before inspecting, resuming, or integrating its runner with OMP or another coordinator.
- omarchy-end4-hybrid-bar: Maintain an Omarchy top bar alongside End-4 Quickshell sidebars and cheat sheet; custom resource modules, outside-click dismissal, wheel scrolling, and runtime verification.
- omarchy-end4-sidebar-dashboard: Maintain detailed End-4 sidebar hardware, Tailscale, and AI allowance widgets, including Codex app-server refresh and QML sequence rendering pitfalls.
- plan-guard: Gate against goal drift and over-engineering; run before writing, approving, delegating, or resuming any plan, brief, feature list, handoff, or 3+-step todo, when new infrastructure, security, services, abstraction, or scale is proposed, and after two failed fixes on one premise. Produces a PASS/RE…
- quickshell-pointer-following-dock: Implement or troubleshoot an End-4 Quickshell dock that follows the pointer across Hyprland monitors without resizing windows, including auto-hide, its searchable app drawer, and runtime verification.
- tesseract-omp-integration: Connect, repair, or diagnose the local Tesseract Obsidian vault through OMP MCP; use for Tesseract setup, broken launchers, indexing, or memory-stack integration.

Descriptions come from the frontmatter of each SKILL.md, punctuation lightly normalized.