Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/perpetualsoftware/pad/claude-mdgit clone --depth 1 https://github.com/PerpetualSoftware/padWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.15292 | $0.15292 |
| Opus 5 | $0.07646 | $0.07646 |
| Sonnet 5 | $0.03058 | $0.03058 |
| Haiku 4.5 | $0.01529 | $0.01529 |
Grade A, and why
pad CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 459 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Pad — Development Guide
What This Is
Pad is a project management tool for developers and AI agents. Single Go binary with embedded SvelteKit web UI, SQLite storage, and multi-agent skill support (Claude Code, Cursor, Windsurf, Codex, OpenCode, Copilot, Amazon Q, Junie).
Related repo: The marketing website (getpad.dev) lives at ../pad-web — a separate SvelteKit site deployed to Vercel.
Architecture
- Backend: Go (cmd/pad/main.go) → REST API (internal/server/) → SQLite (internal/store/)
- Frontend: SvelteKit 2 + Svelte 5 (web/src/) → static build embedded in Go binary
- Data model: Workspaces → Collections (typed with JSON schemas) → Items (structured fields + rich content)
- CLI: Cobra commands in cmd/pad/main.go, HTTP client in internal/cli/
- Agent skill: Single natural-language
/padskill in skills/pad/SKILL.md
Build & Install
make build # Build web UI + Go binary (./pad)
make install # Build, kill server, install to ~/.local/bin/pad, restart
make build-go # Build Go only (skip web — faster when only backend changes)
make test # Run Go tests
make web # Build web UI only
make dev-web # Run SvelteKit dev server (hot reload on :5173)
After making changes, always run make install to rebuild the binary, install it, and restart the server. The web UI at http://localhost:7777 will reflect the changes.
Quick iteration loop
- Backend only:
make install(skips web rebuild if no frontend changes — edit Makefile to usebuild-goinstead ofbuildin the install target) - Frontend only:
make web && make installor usemake dev-webfor hot reload during development - Full rebuild:
make install
Working in a git worktree
Agent sessions take a git worktree per task rather than sharing the main checkout (a shared checkout means a shared stash stack, branch state, and dirty files across sessions). Three rules keep web tooling working there:
- Symlinking
web/node_modulesto the main checkout's copy is fine (and fast). Vitest,vite build, andnpm run checkall work through the symlink. - A fresh worktree has no
web/.svelte-kit(gitignored, generated). Runnpx svelte-kit syncinweb/before any vitest/vite command — ornpm run check, which syncs first. Without it, vitest fails withFailed to load tsconfig '.svelte-kit/tsconfig.json': Tsconfig not foundregardless of hownode_moduleswas set up. (This missing generated dir was historically misdiagnosed as a symlink problem —npm ci"fixed" it only because itspreparescript runssvelte-kit sync.) - Never run
npm ciin a worktree whoseweb/node_modulesis a symlink — including via make.npm cilives in thewebtarget, so every target whose dependency chain reaches it is off-limits too: currentlyweb,build,install,serve,web-check, andcheck(viaweb-check). Everything else —build-go,dev,restart,test,test-pg,lint,vuln,web-test,dev-web,clean— never reachesnpm ci.npm cideletes through the symlink into the shared tree, breaking every other worktree and session at once with a confusingvitest: not found. If you want a real, isolatednode_modulesinstead of a symlink,npm ciin an un-symlinkedweb/is ~5s on a warm cache and regenerates.svelte-kitas a side effect.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 459 lines · 15,292 tokens per session scan A c48ca81d7fc8
pad CLAUDE.md is an instructions file published in the GitHub repository PerpetualSoftware/pad (146 stars, last pushed 3d ago), licensed Apache-2.0. It adds 15,292 tokens to every session, about $0.0765 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
sverklo CLAUDE.md
Instructions for sverklo/sverklo, covering sverklo development, using sverklo tools, project structure and build & test.
brain-os AGENTS.md
Instructions for brainOS-HQ/brain-os, covering agents.md, context, the hard rule, tool routing (in order) and mutation safety.
brain-os copilot-instructions.md
Instructions for brainOS-HQ/brain-os, covering github copilot instructions, read this first, critical rules (also in agents.md) and known bugs (v0.4.1 fixes pending).
brain-os CLAUDE.md
Instructions for brainOS-HQ/brain-os, covering claude.md, agent instructions, build / dev, publish flow and project memory.
asynkor CLAUDE.md
Instructions for asynkor/asynkor, covering asynkor — team coordination, 1. connect — read the team brain, 2. first-time setup — populate long-term context, 3. start work — declare intent + acquire leases and 4. during work — leases protect your files.
tuskledger-mcp AGENTS.md
AGENTS.md instructions for BradMorphsters/tuskledger-mcp, covering agents.md, what this project is, permission boundaries, safe to do without asking and ask the user first.