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/punt-labs/punt-kit/agents-mdgit clone --depth 1 https://github.com/punt-labs/punt-kitWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/punt-labs/punt-kit/agents-md)<a href="https://agentmods.dev/instructions/punt-labs/punt-kit/agents-md"><img src="https://agentmods.dev/badge/instructions/punt-labs/punt-kit/agents-md.svg" alt="Measured on agentmods" height="20"></a>What 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 | $0.03019 | $0.03019 |
| Opus 5 | $0.01510 | $0.01510 |
| Sonnet 5 | $0.00604 | $0.00604 |
| Haiku 4.5 | $0.00302 | $0.00302 |
Grade A, and why
punt-kit AGENTS.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 4d 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 — 168 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent Instructions
This is punt-kit — the standards, design patterns, and scaffolding repo for Punt Labs. When working in any Punt Labs project, consult these standards and patterns.
Beads
This project uses bd (beads) for issue tracking. Run bd onboard to get started.
bd ready # Find available work
bd show <id> # View issue details
bd update <id> --status in_progress # Claim work
bd close <id> # Complete work
bd sync # Sync with git
Standards
Each standard lives in its own document under standards/.
| Standard | What it covers |
|---|---|
| Distribution | Install paths (marketplace, PyPI, .mcpb), dependency pinning, installation scope, init vs install, uninstall requirements |
| CLI | CLI + plugin duality, required subcommands (install, doctor, version, serve), --json flag, shell completion |
| Tool Enable/Disable | User-owned CLAUDE.md, the bare @-import line, the zoned .punt-labs/<tool>/ subtree (vendored zone tool-owned), the enabled marker, enable/disable verbs, audit checks, sentinel migration |
| Plugins | plugin.json, MCP server declaration, extension points, required hooks (SessionStart + PostToolUse), tool naming, gitignore checklist |
| Hooks | Claude Code state machine, three-layer dispatch, shell gates, Python handlers, decision-block and workflow-gate patterns, startup performance, audit checklist |
| Permissions | Allow/prompt/deny tiers, settings.json vs settings.local.json, required allow and deny rules, plugin-distributed permissions |
| Filesystem | ~/.punt-labs/<tool>/ home layout, reserved subdirectories, per-project activation marker, migration from legacy dot-directories |
| Repo-Local State Directory | <repo>/.punt-labs/<tool>/ tool roots plus the machine-local .punt-labs/local/ zone, committed except local-convention paths (local/, *.local, *.local.*), live state never tracked (the seal pattern, chunk-based), the canonical gitignore block, subtree zones (vendored/config/local/marker) |
| Integration | Peer-tool discovery tiers L0–L5, presence markers, graceful degradation, integration matrix, bundled integrations |
| Logging | Centralized dictConfig, log locations, atomic append, file permissions, content escaping |
| Naming | Repo names, PyPI names, CLI names, slash commands, versioning (semver) |
| GitHub | Branch protection, CI/CD workflows, Copilot code review, required status checks, repo settings |
| Workflow | Issue tracking, workflow tiers, branch discipline, commits, quality gates, code review, session close protocol, design decision logs, cross-project integration |
| PR and Review | Local-first review: the PR as merge gate, review agents, remote reviewers as second opinion |
| Git | Git mechanics: branch check before commit, no-rewrite of pushed branches, conflict resolution, post-merge cleanup ordering, worktree lifecycle, submodules, mission trailers, stop-and-ask |
| Release Process | The /punt:auto release playbook, deterministic phases 1–11, cross-repo propagation |
| Release Requirements | End-state artifacts a release must satisfy in the originating repo and siblings |
| README | Required README structure, install trust tiers, tone and positioning |
| Makefile | Required Makefile targets; make check as the quality-gate entry point |
| Shell | POSIX sh vs bash, strict mode, shellcheck, install-script rules |
| Agent Engineering | Operating defaults for AI coding agents: judgment, scope, verification |
| Architecture | The engine-and-clients model: one engine, thin library/CLI/MCP/REST client surfaces |
| OO | The object-oriented stance shared by Python, Swift, and Pharo |
| Python | uv, ruff, mypy, pyright, pytest, typer + rich, FastMCP, the OO ratchet, release workflow |
| Go | Idiomatic Go: modules, small interfaces, explicit errors, go vet + staticcheck |
| C | Idiomatic C: memory ownership, -Werror, sanitizers |
| Swift | Idiomatic + protocol-oriented Swift; swiftformat, swiftlint, XcodeGen, XCTest |
| Pharo | Pharo/Smalltalk, image-based development, in-image linting |
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.
- 4d ago First seen · 168 lines · 3,019 tokens per session scan A 5d7adf6d7546
punt-kit AGENTS.md is an instructions file published in the GitHub repository punt-labs/punt-kit (1 stars, last pushed yesterday), licensed MIT. It adds 3,019 tokens to every session, about $0.0151 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 2026-08-31.
Other instructions, from other repositories
prfaq CLAUDE.md
Instructions for punt-labs/prfaq, covering how i write code, mandatory reading, no "pre-existing" excuse, standards and development workflow.
z-spec CLAUDE.md
Claude Code instructions for punt-labs/z-spec, covering z-spec, mandatory reading, read this first, architecture and key architectural boundary: commands vs. surfaces.
quarry CLAUDE.md
Instructions for punt-labs/quarry, covering quarry, mandatory reading, architecture, how a query works and how ingestion works.
vox CLAUDE.md
Instructions for punt-labs/vox, covering vox, mandatory reading, architecture, how synthesis works and key architectural boundary: daemon vs. clients.
biff CLAUDE.md
Claude Code instructions for punt-labs/biff, covering how i write code, no "pre-existing" excuse, no git submodules, standards and testing pyramid.
beadle CLAUDE.md
Instructions for punt-labs/beadle, covering beadle, what beadle is, identity, no "pre-existing" excuse and standards.