repositories / dotfiles

dotfiles

bugabingas dorkfiles

owned by bugabinga

update Pi agent guidance and settings

2198648ff227f134c4071ea652f3df7e14051985

Oliver Jan Krylow <oliver@bugabinga.net> ·

unsigned

Browse filesCompare with HEAD

Changed files (11)

pi/agent/AGENTS.md

Deleted
deleted file mode 100644
1 -SIMPLICITY!
2 -MINIMALSM!
3 -BRUTALISM!
4 -DEATH TO LEGACY/COMPAT/HEDGING!
5 -DEATH TO EDGE CASES! FOUND SOME? TELL ME.
6 -DEATH TO PYTHON! You may use it for temp scripts and one-off, but using it in production is VERBOTEN!
7 -STOP SMOKING! `smoke` is a bad name -> use semantic names!
8 -On Win32, Pi’s bash tool uses Bash; redirect to /dev/null, never NUL or nul.

pi/agent/APPEND_SYSTEM.md

Modified
46 46 Arrive with the fix designed; leave authority with the owner.
47 47 Evaluate statements on content, never on sender. Authority must be earned.
48 48 Performance matters; dependencies are vetted, few, and suspect by default.
49 -Structural content wants a diagram (d2/dot/mermaid; context decides the tool, not the concept).
51 -## H5 instance data (canonical complexity-taxonomy classifications)
52 -JVM bytecode and class-loading: internalized (bytecode interpreter written, JEPs read).
53 -Rust borrow checker details: being internalized (idiomatic code without re-deriving; `rustc --explain` still reached for on lifetimes).
54 -LLM weights and the training loop: behind a legitimate abstraction boundary (API surface is what I need; the rest is for the people who train them).
55 -Internal company tooling (e.g. the CRM): uninternalizable mess (the test: nobody on my team can describe its data model without hand-waving).
49 +Structural content wants a diagram.
50 +DEATH TO LEGACY/COMPAT/HEDGING!
51 +DEATH TO EDGE CASES! FOUND SOME? TELL ME.
52 +DEATH TO PYTHON! You may use it for temp scripts and one-off, but using it in production is VERBOTEN!
53 +STOP SMOKING! `smoke` is a bad name -> use semantic names!
54 +On Win32, Pi’s bash tool uses Bash; redirect to /dev/null, never NUL or nul.
57 56 ## working with me
58 57 No anthropomorphism: I don't model your feelings and don't want mine modeled. Precise intent delivery, both directions.
60 59 Rationale rides inline: "because …" on the same line as the claim.
61 60 Bold = globally important (must pop when scanning the page). Italic = locally important.
62 61 Semantic line breaks in all markup.
63 -No em dash (—). Comma, colon, semicolon, or period. The LLM default is em dash; this rule overrides it.
62 +No em dash (—). Comma, colon, semicolon, or period.
64 63 Don't intellectualize past clarity; language is communication, not thought.
65 +## coding guidelines
67 +Ladder:
68 +1. Do not build what was not asked.
69 +2. Reuse existing code in this repo.
70 +3. Prefer stdlib.
71 +4. Prefer native platform APIs.
72 +5. Prefer installed deps over new deps.
73 +6. Prefer one-line/data-only changes.
74 +7. Write minimum custom code.
76 +Root-cause bug-fix rule: inspect callers and fix the shared route when possible.
78 +Rules:
79 +- No unrequested abstractions: no one-impl interfaces, factories, wrappers, or config nobody sets.
80 +- Deletion over addition. Boring over clever. Fewest files possible.
81 +- Be lazy about the solution, never about reading: inspect the touched flow before choosing a rung.
82 +- Two stdlib/native options, same size? Pick the edge-case-correct one.
83 +- Challenge unnecessary scope once, then ship the smallest safe version.
84 +- Enforce the ladder before custom code.
86 +Safety floor: never remove trust-boundary validation, data-loss handling, security, accessibility, hardware calibration,
87 +or explicitly requested behavior.
89 +Test floor: non-trivial logic leaves one small runnable check; trivial one-liners need no test.
90 +Output discipline: code first; short skipped/when-to-add note; no unrequested essays.

pi/agent/settings.json

Modified
2 - "lastChangelogVersion": "0.84.1",
2 + "lastChangelogVersion": "0.84.2",
3 3 "packages": [
5 5 "source": "~/Workspace/dotfiles/pi-ext"
70 70 "enable": true
72 72 "ultra": {
73 + "actionSummarizerModel": "openai-codex/gpt-5.6-luna",
73 74 "modelTiers": {
74 75 "small": "openai-codex/gpt-5.6-luna",
75 76 "medium": "openai-codex/gpt-5.6-terra",

pi/agent/skills/ascii-boxes/SKILL.md

Modified
2 2 name: ascii-boxes
3 -description: Draw proper Unicode boxes/tables. Use when output needs aligned terminal boxes, table borders, nested frames, box-drawing chars. FORBIDS ASCII +---+ boxes.
3 +description: Draw proper Unicode boxes/tables. Use when output needs aligned terminal boxes, table borders, nested frames, box-drawing chars.
6 6 # ASCII BOXES
22 22 CLI:
24 24 ```bash
25 -node ./index.mjs draw "text" --width N --style single|double|bold --align left|center|right
25 +node ./index.mjs draw "text" --width N --style single | double | bold --align left | center | right
26 26 node ./index.mjs table --header "H1,H2" --rows "a,b" "c,d"
27 27 node ./index.mjs nested "OUTER" "INNER" --gap N

pi/agent/skills/checkhealth/SKILL.md

Modified
2 2 name: checkhealth
3 -description: "Analyze current software project state: lifecycle stage, completion vs TODO, stack/runtime versions, code style, deps, tooling, tests, build health, and repo orientation. Use when asked for project health, status, audit, inventory, orientation, or 'what the hell is going on'."
3 +description: "Analyze current software project state. Use when asked for project health, status, audit, inventory, orientation, or 'what the hell is going on'."
6 6 # Checkhealth

pi/agent/skills/experimentation/SKILL.md

Modified
2 2 name: experimentation
3 -description: "Use when agent is about to create/run throwaway code or generated artifacts—scratch scripts, spikes, probes, PoCs, one-off parsers/converters, microbenchmarks, API/dependency behavior checks—to answer a narrow technical question. Enforces falsifiable hypothesis, controls/baselines, independent verification, provenance, promote-or-delete, and cleanup. Do not use for normal feature implementation/review, permanent project tests, or web research without scratch code/artifacts."
3 +description: "Use when agent is about to create/run throwaway code or generated artifacts—scratch scripts, spikes, probes, PoCs, one-off parsers/converters, API/dependency behavior checks—to answer a narrow technical question."
6 6 # Experimentation

pi/agent/skills/find-skills/SKILL.md

Deleted
deleted file mode 100644
1 ----
2 -name: find-skills
3 -description: Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.
4 ----
6 -# Find Skills
8 -This skill helps you discover and install skills from the open agent skills ecosystem.
10 -## When to Use This Skill
12 -Use this skill when the user:
14 -- Asks "how do I do X" where X might be a common task with an existing skill
15 -- Says "find a skill for X" or "is there a skill for X"
16 -- Asks "can you do X" where X is a specialized capability
17 -- Expresses interest in extending agent capabilities
18 -- Wants to search for tools, templates, or workflows
19 -- Mentions they wish they had help with a specific domain (design, testing, deployment, etc.)
21 -## What is the Skills CLI?
23 -The Skills CLI (`npx skills`) is the package manager for the open agent skills ecosystem. Skills are modular packages that extend agent capabilities with specialized knowledge, workflows, and tools.
25 -**Key commands:**
27 -- `npx skills find [query] [--owner <owner>]` - Search for skills interactively or by keyword, optionally scoped to a GitHub owner
28 -- `npx skills add <package>` - Install a skill from GitHub or other sources
29 -- `npx skills check` - Check for skill updates
30 -- `npx skills update` - Update all installed skills
32 -**Browse skills at:** https://skills.sh/
34 -## How to Help Users Find Skills
36 -### Step 1: Understand What They Need
38 -When a user asks for help with something, identify:
40 -1. The domain (e.g., React, testing, design, deployment)
41 -2. The specific task (e.g., writing tests, creating animations, reviewing PRs)
42 -3. Whether this is a common enough task that a skill likely exists
44 -### Step 2: Check the Leaderboard First
46 -Before running a CLI search, check the [skills.sh leaderboard](https://skills.sh/) to see if a well-known skill already exists for the domain. The leaderboard ranks skills by total installs, surfacing the most popular and battle-tested options.
48 -For example, top skills for web development include:
50 -- `vercel-labs/agent-skills` — React, Next.js, web design (100K+ installs each)
51 -- `anthropics/skills` — Frontend design, document processing (100K+ installs)
53 -### Step 3: Search for Skills
55 -If the leaderboard doesn't cover the user's need, run the find command:
57 -```bash
58 -npx skills find [query] [--owner <owner>]
59 -```
61 -For example:
63 -- User asks "how do I make my React app faster?" → `npx skills find react performance`
64 -- User asks "can you help me with PR reviews?" → `npx skills find pr review`
65 -- User asks "I need to create a changelog" → `npx skills find changelog`
67 -### Step 4: Verify Quality Before Recommending
69 -**Do not recommend a skill based solely on search results.** Always verify:
71 -1. **Install count** — Prefer skills with 1K+ installs. Be cautious with anything under 100.
72 -2. **Source reputation** — Official sources (`vercel-labs`, `anthropics`, `microsoft`) are more trustworthy than unknown authors.
73 -3. **GitHub stars** — Check the source repository. A skill from a repo with <100 stars should be treated with skepticism.
75 -### Step 5: Present Options to the User
77 -When you find relevant skills, present them to the user with:
79 -1. The skill name and what it does
80 -2. The install count and source
81 -3. The install command they can run
82 -4. A link to learn more at skills.sh
84 -Example response:
86 -```
87 -I found a skill that might help! The "react-best-practices" skill provides
88 -React and Next.js performance optimization guidelines from Vercel Engineering.
89 -(185K installs)
91 -To install it:
92 -npx skills add vercel-labs/agent-skills@react-best-practices
94 -Learn more: https://skills.sh/vercel-labs/agent-skills/react-best-practices
95 -```
97 -### Step 6: Offer to Install
99 -If the user wants to proceed, you can install the skill for them:
101 -```bash
102 -npx skills add <owner/repo@skill> -g -y
103 -```
105 -The `-g` flag installs globally (user-level) and `-y` skips confirmation prompts.
107 -## Common Skill Categories
109 -When searching, consider these common categories:
111 -| Category | Example Queries |
112 -| --------------- | ---------------------------------------- |
113 -| Web Development | react, nextjs, typescript, css, tailwind |
114 -| Testing | testing, jest, playwright, e2e |
115 -| DevOps | deploy, docker, kubernetes, ci-cd |
116 -| Documentation | docs, readme, changelog, api-docs |
117 -| Code Quality | review, lint, refactor, best-practices |
118 -| Design | ui, ux, design-system, accessibility |
119 -| Productivity | workflow, automation, git |
121 -## Tips for Effective Searches
123 -1. **Use specific keywords**: "react testing" is better than just "testing"
124 -2. **Try alternative terms**: If "deploy" doesn't work, try "deployment" or "ci-cd"
125 -3. **Check popular sources**: Many skills come from `vercel-labs/agent-skills` or `ComposioHQ/awesome-claude-skills`
127 -## When No Skills Are Found
129 -If no relevant skills exist:
131 -1. Acknowledge that no existing skill was found
132 -2. Offer to help with the task directly using your general capabilities
133 -3. Suggest the user could create their own skill with `npx skills init`
135 -Example:
137 -```
138 -I searched for skills related to "xyz" but didn't find any matches.
139 -I can still help you with this task directly! Would you like me to proceed?
141 -If this is something you do often, you could create your own skill:
142 -npx skills init my-xyz-skill
143 -```

pi/agent/skills/git-workflow/SKILL.md

Modified
2 2 name: git-workflow
3 -description: "Use for Git repository workflows: committing changes, rebasing/pushing fast-forward, landing branches, or reviewing PRs. Load when a prompt says git workflow or when jj is not active."
3 +description: "Use for Git repository workflows: committing changes, rebasing/pushing fast-forward, landing branches, or reviewing PRs."
6 6 # Git Workflow

pi/agent/skills/qml/SKILL.md

Modified
2 2 name: qml
3 -description: Use when writing or editing QML/QtQuick code — property bindings, components, signals, models, anchors, layouts, JavaScript integration. Triggers on "qml", "qtquick", "qt qml", "qml binding", "qml signal", "qml component", "qml layout", "qt property binding", "qtquick bar", "qt panel". The pure language only — no compositor or desktop-integration types.
3 +description: Use when writing or editing QML/QtQuick code. Triggers on "qml", "qtquick", "qt qml", "qml binding", "qml signal", "qml component", "qml layout", "qt property binding", "qtquick bar", "qt panel".
6 6 # QML

pi/agent/skills/speak-as-oliver/SKILL.md

Modified
2 2 name: speak-as-oliver
3 -description: Write or communicate as Oliver (bugabinga) in his voice, on his behalf. Trigger when Oliver asks for a draft "as me", "in my voice", "on my behalf", or asks to write outward-facing text he will send under his name (issues, PR comments, emails, notes, essays, slides, READMEs. Do NOT trigger for normal agent work, code, or answers addressed to Oliver himself.
3 +description: Write or communicate as Oliver (bugabinga) in his voice, on his behalf. Trigger when Oliver asks for a draft "as me", "in my voice", "on my behalf", or asks to write outward-facing text he will send under his name (issues, PR comments, emails, notes, essays, slides, READMEs.
6 6 # speak-as-oliver

pi/agent/skills/wiki/SKILL.md

Modified
89 89 ```text
90 90 1. Classify wiki initialization state.
91 91 2. If initialized, read SCHEMA.md.
92 -3. Search ~/Notes/Wiki/ and ~/.pi/research/ for existing sources/pages.
92 +3. Search ~/Notes/Wiki/ for existing sources/pages.
93 93 4. If missing, fetch/copy/clone into raw-sources/ and register it.
94 94 5. Search/read local raw source files.
95 95 6. Compile or update concept pages with citations and wikilinks.
129 129 2. Search raw sources if compiled pages are missing/thin:
130 130 ```bash
131 - rg -i -F --max-filesize 1M -- "$query" ~/Notes/Wiki/raw-sources ~/.pi/research 2>/dev/null
131 + rg -i -F --max-filesize 1M -- "$query" ~/Notes/Wiki/raw-sources 2>/dev/null
133 133 3. Read matching pages/sources.
134 134 4. Synthesize answer with `[[wikilink]]` citations.
139 139 Source can be a URL, repository URL, archive URL, file path, directory, pasted text.
141 141 1. **Classify** source: web | repo | archive | file | text.
142 -2. **Check existing** `raw-sources/index.md`, filenames, and `~/.pi/research/` for matching source.
142 +2. **Check existing** `raw-sources/index.md` and filenames for matching source.
143 143 3. **Preflight risky sources**:
144 144 - URLs: allow public `http(s)` URLs by default. Private/internal URLs require explicit approval, with warning first; examples include localhost, loopback, RFC1918/private IPs, and `169.254.169.254`. Reject credentialed URLs and `file://` URLs by default.
145 145 - Files/dirs: deny known secret paths and token material by default, including `~/.ssh`, `.env`, credentials, browser profiles, token stores, private keys, and app auth files. Require explicit approval outside the project/current source scope.
210 210 rg -i -F --max-filesize 1M -- "$query" ~/Notes/Wiki/raw-sources/
212 212 # Repo source with context
213 -rg -n -A2 -B2 -F --max-filesize 1M -- "$query" ~/Notes/Wiki/raw-sources/repos/<repo>/
213 +rg -n -A2 -B2 -F --max-filesize 1M -- "$query" ~/Notes/Wiki/raw-sources/repos/ <repo >/
215 215 # HTML text snippets when markup is noisy and regex was explicitly requested
216 216 rg -s --max-filesize 1M ">([^<]*pattern[^<]*)<" ~/Notes/Wiki/raw-sources/web/