pi coding agent skill

app/priv/blog/engineering/2026/07-16-at-coding-agent-skill-to-guide-editing-in-place.md

@@ -31,7 +31,13 @@ My users use the application through screens, I can model the domain as much as
 
 Doing that took extending tests to output HTML in the right places. For this Pi + Qwen worked quite well, in the get it Run sense. Get it right, less so. I had some ideas on extracting a DSL, broke them down into tasks. When I do that, I tend to over-engineer. My coding agent less so, but once it worked, I had 400 lines of code for a test that was just spanning three screens. 
 
-This took 1.6 minutes, which is not particularly fast. But I was doing other things. There is refactoring tooling for Elixir, but it is not that easy to setup, I am in the process of doing it. I have paid attention to the extract setup refactoring more closely and will do the next one by hand...
+Refactoring by hand and by Agent
+----
+
+This took 1.6 minutes, which is not particularly fast. But I was doing other things. There is refactoring tooling for Elixir, but it is not that easy to setup, I am in the process of doing it. I have paid attention to the extract setup refactoring more closely and will do the next one by hand, probably. I have done some extract functions by hand to keep a feel for Elixir, and also because it sometimes takes less time to just do it than to explain it.
+
+Aside: when the coding agent reports the number of tests
+----
 
 Also note, that as Pi / Qwen reported success, that not all the tests have been run (there are 998 tests in total, this is an umbrella project with `make test` at the root)
 
@@ -39,11 +45,31 @@ Also note, that as Pi / Qwen reported success, that not all the tests have been
 > - ✅ Clean compilation (no warnings)
 > - ✅ Committed as 2e2086d
 
+The rubber duck in the blog post is strong. All the tests are run, but in three rounds, as this is a small mono-repo. I am attempting to keep some things that are not specific for the domain, like flexible forms, separate, without spinning them off in to separate repos. The coding agent picked up on the last run, which has 739 tests without mentioning the other ones. 
+
+I was tempted to blame the agent for not running all tests, but they are part of a post-commit hook, so they get run deterministically, no matter what.
+
 About the skill
 ====
 
 It is probably overkill to use `git status --porcelain` a search through all files might work as well, or better. `rg` is not working in my agents' sandbox for some reason, I need to figure this out still. It is also specific to elixir in places. Also not sure if prioritization is necesary. I put it in, because when I see multiple things that need work in a piece of code, I want to annotate before I forget (this is analogous to putting test lists in code, but for refactorings too).
 
+I had expected to get this to work in one prompt, but maybe my perfectionism got the better of me, once again. I thought do do this with sub-agents in one go (looking at the skill) and getting the context picked up besides my one line instruction, was interesting. See the *Context Resolution* section below. 
+
+Note that none of the `@pi` instructions will make it into the commit history, if you work one instruction at a time. Luckily the skill has an example:
+
+```elixir
+# @Pi: move to Swimex.Report
+defp extract_fragment(full_html, selector) do
+```
+
+That already needed  a re-prompt, because I had specified 'pi' in the instruction to create the skill, and used 'Pi' when trying it out. Moving a function to another module is simple, but annoying: 
+
+- Copy function over to new module.
+- Import function in old module
+- Make sure it is referenced correctly by all call sites
+- Delete original function
+
 The At Pi skill in full
 =====