pikers
/
piker
4
1
Fork 0
Code Issues 14 Pull Requests 12 Packages Projects Releases Wiki Activity

Add .claude/skills/* files from gap-annotator perf sesh with ma boi #69

Open
goodboy wants to merge 1 commits from claudy_skillz into hist_backfill_fixes
pull from: claudy_skillz
merge into: pikers:hist_backfill_fixes
Conversation 0 Commits 1 Files Changed 4
goodboy commented 2026-01-31 20:36:18 +00:00
Owner

Eventually going to rebase this onto main obvi but put it up right away so it can be landed quickly and ideally i can continue ongoing dev work with it mainlined.

All the content comes from a very good prompt-eng session i had this week using claude to optimize a newly revamped time-series gap detector i originally started realizing we needed back during the nativedb work started on GH (see historical details outline in #62).

This inspired me to start now working toward the idealism that we can make working with all AI tools not only seamless but alot of fun for all (potential new) collabs, both newb and core-hacker-degen alike!

Core principles and long run goals

From my PoV specifically, i want to build a persona/character/experience-set for coding aids/“assistants” which,

  • reflects the interaction of paired programming with a real senior core dev; the entire feel should be like you’re asking for help in one of our chats.

  • emphasizes a project focused capability set and knowledge base grounded in our core engineering principles:

    • conc/queueing/distributed-systems dedicated to hyper-correctness, resiliency and fault tolerance (despite chaotic mkt conditions), poor-performing-code = invalid code, a decentralized-systems-first design.
    • a traders-first mentality which dictates the platform should never (or at the least hyper-minimize-for) failures that can lose the user $$ during live trades.
    • a scientfic-data-anal/numerics-expert underpinning which drives the design of all supporting and overlying subsystems and ensures low-latency, real-time operation within a distributed and collaborative env; open, performant data sharing/processing/storage/mining is our default use case.

  • have a very non-corporate “vibe” in all aspects; coding on piker should be fun and chill, something you want to work on for yourself not something you’re getting paid to work on by some overlord..

Impl deats: groking “skills files” with claude

The main idea with these (i guess “config”) files is to start maintaining and extending the state of the AI interaction experience.

See the anthropic docs,

Of note for me was,

  • wrt "output styles, being able to have more/less terse/technical feedback depending on your current workflow context:

    • while trading we should optimize for noise/distraction-reduction, thinking/reacting fast with zero mental errors.
    • while hacking/designing we should emphasize tinkering/researching/profiling and generally having fun with ideas generation and naming.
    • while backtesting / testing (auto) strats we should try to be rigorous in our open mindedness and creativity, old ideas aren’t bad but they likely have stopped working (at least at some point) so we should always be mindful of our preset assumptions.

  • wrt “subagents” building out a set for the bullets listed in the above “core principles” section.

  • wrt “mcp”, creating services for monitoring/configuring/debugging/coordinating-&-remotely-controlling various actor-daemons in our runtime to enable traders and/or hilevel anal/research users from requiring deep engineering knowledge to manage their (distributed) environment.

I also need to read the docs more thoroughly so likely more to come in follow up as well ;)

ToDos for this to land

  • the current set of files generated by claude when i asked seem to not adhere to the above docs? particularly,

    so i think i’ll start by trying to get claude to read its own docs and adjust XD

  • there are various tweaks to be made in the skills file’s content themselves,

    • adjustments/corrections to terms/emojis/syntaxes in the slang guide
    • read through the other technical files and extend this list as needed.
Eventually going to rebase this onto `main` obvi but put it up right away so it can be landed quickly and ideally i can continue ongoing dev work with it mainlined. All the content comes from a very good prompt-eng session i had this week using `claude` to optimize a newly revamped time-series gap detector i originally started realizing we needed back during the `nativedb` work started on GH (see historical details outline in https://pikers.dev/pikers/piker/pulls/62). This inspired me to start **now** working toward the idealism that we can make working with all AI tools not only seamless but alot of fun for all (potential new) collabs, both newb and core-hacker-degen alike! --- ## Core principles and long run goals From my PoV specifically, i want to build a persona/character/experience-set for coding aids/"assistants" which, - reflects the interaction of paired programming with a real senior core dev; the entire feel should be like you're asking for help in one of our chats. - emphasizes a project focused *capability set* and *knowledge base* grounded in our core engineering principles: + conc/queueing/distributed-systems dedicated to hyper-correctness, resiliency and fault tolerance (despite chaotic mkt conditions), poor-performing-code = *invalid code*, a *decentralized-systems-first* design. + a traders-first mentality which **dictates** the platform should never (or at the least hyper-minimize-for) failures that can lose the user $$ during live trades. + a scientfic-data-anal/numerics-expert underpinning which drives the design of all supporting and overlying subsystems and ensures low-latency, real-time operation within a distributed and collaborative env; open, performant data sharing/processing/storage/mining is our **default use case**. - have a **very** non-corporate "vibe" in all aspects; coding on `piker` should be *fun and chill*, something *you want to work on for **yourself*** not something *you're getting paid to work on by some overlord*.. --- ## Impl deats: groking "skills files" with `claude` The main idea with these (i guess "config") files is to start maintaining and extending the state of the AI interaction experience. See the anthropic docs, - https://code.claude.com/docs/en/skills - interaction vibez: https://code.claude.com/docs/en/output-styles - env-based automation: https://code.claude.com/docs/en/hooks-guide - subagents: https://code.claude.com/docs/en/sub-agents - mcp: https://code.claude.com/docs/en/mcp Of note for me was, - wrt "output styles, being able to have more/less terse/technical feedback depending on your current workflow context: + while trading we should optimize for noise/distraction-reduction, thinking/reacting fast with zero mental errors. + while hacking/designing we should emphasize tinkering/researching/profiling and generally having fun with ideas generation and naming. + while backtesting / testing (auto) strats we should try to be rigorous in our open mindedness and creativity, old ideas aren't bad but they likely have stopped working (at least at some point) so we should always be mindful of our preset assumptions. - wrt "subagents" building out a set for the bullets listed in the above "core principles" section. - wrt "mcp", creating services for monitoring/configuring/debugging/coordinating-&-remotely-controlling various actor-daemons in our runtime to enable traders and/or hilevel anal/research users from requiring deep engineering knowledge to manage their (distributed) environment. I also need to read the docs more thoroughly so likely more to come in follow up as well ;) --- ## ToDos for this to land - [ ] the current set of files generated by `claude` when i asked seem to not adhere to the above docs? particularly, + https://code.claude.com/docs/en/skills#frontmatter-reference is not present? + there is no `SKILLS.md` in the subdir but [the docs say it's required?](https://code.claude.com/docs/en/skills#configure-skills) + ensuring content is organized accordingly per the [types of skill](https://code.claude.com/docs/en/skills#types-of-skill-content) section? so i think i'll start by trying to get `claude` to read its own docs and adjust XD - [ ] there are various tweaks to be made in the skills file's content themselves, + [ ] adjustments/corrections to terms/emojis/syntaxes in the [slang guide](https://www.pikers.dev/pikers/piker/pulls/69/files#diff-03f32be9439ab086216772d2c15005df1762de44) + [ ] read through the other technical files and extend this list as needed.
goodboy added 1 commit 2026-01-31 20:36:18 +00:00
This pull request can be merged automatically.
This branch is out-of-date with the base branch
You are not authorized to merge this pull request.
You can also view command line instructions.

Step 1:

From your project repository, check out a new branch and test the changes.
git checkout -b claudy_skillz hist_backfill_fixes
git pull origin claudy_skillz

Step 2:

Merge the changes and update on Gitea.
git checkout hist_backfill_fixes
git merge --no-ff claudy_skillz
git push origin hist_backfill_fixes
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No Label
Milestone
Clear milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
No Assignees
1 Participants
Notifications
Due Date
The due date is invalid or out of range. Please use the format 'yyyy-mm-dd'.

No due date set.

Dependencies

No dependencies set.

Reference: pikers/piker#69
There is no content yet.
Delete Branch "claudy_skillz"

Deleting a branch is permanent. Although the deleted branch may exist for a short time before cleaning up, in most cases it CANNOT be undone. Continue?