1  #+title: Pi-AgentStack
 2  #+author: HaQadosch
 3  #+date: [2026-04-22 Wed]
 4  #+startup: indent
 5  
 6  * Purpose
 7  
 8  Research and evaluation workspace for the /agentic stack/ pattern
 9  introduced by [[https://github.com/codejunkie99/agentic-stack][codejunkie99/agentic-stack]] -- a portable =.agent/= brain
10  (memory, skills, protocols) that mounts into any coding-agent harness
11  (Claude Code, Cursor, OpenCode, Windsurf, standalone Python, ...).
12  
13  This directory is *not* an implementation. It is a study workspace
14  used to decide whether the pattern is the right fit before committing
15  to adoption. If/when adopted, the actual =.agent/= install will live
16  elsewhere -- likely mounted into =PiCodingAgent/=, or as a standalone
17  portable brain that multiple =Pi-*= projects share.
18  
19  * How to Use This Workspace
20  
21  - [[file:upstream-walkthrough.org][upstream-walkthrough.org]] :: Analysis of the reference repo. Read
22    first to understand what is actually on offer. Treat as
23    frozen-in-time notes -- re-fetch from source if the repo moves.
24  - [[file:evaluation.org][evaluation.org]] :: Fit criteria and running assessment. This is
25    where judgement accumulates. Edit freely as the evaluation
26    progresses.
27  - [[file:decisions.org][decisions.org]] :: ADR-style log of committed decisions. Keep
28    small. Nothing goes here until a decision is actually made, with a
29    date, a rationale, and a reversal condition.
30  
31  * Scope Boundary
32  
33  Do not create =.agent/= directories, mount scripts, or skill stubs in
34  this workspace. Those belong in the target project /once/ the adopt
35  decision lands. Mixing study and implementation here defeats the
36  purpose of the separation -- the moment you have a half-built
37  =.agent/= here, the evaluation is biased toward "yes, adopt" by sunk
38  cost.