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/kubetail-org/kstack/claude-mdgit clone --depth 1 https://github.com/kubetail-org/kstackWrote 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/kubetail-org/kstack/claude-md)<a href="https://agentmods.dev/instructions/kubetail-org/kstack/claude-md"><img src="https://agentmods.dev/badge/instructions/kubetail-org/kstack/claude-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 | $0.04033 | $0.04033 |
| Opus 5 | $0.02017 | $0.02017 |
| Sonnet 5 | $0.00807 | $0.00807 |
| Haiku 4.5 | $0.00403 | $0.00403 |
Grade C, and why
kstack CLAUDE.md scanned grade C with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Downloads and executes remote codehighSupply chain
curl | sh runs whatever the server returns today, which is not necessarily what it returned when this was reviewed.
- `scripts/bootstrap.sh` — the hosted `curl | bash` getter. Clones an upstream checkout, then execs `$UPSTREAM/scripts/install`. Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
- `scripts/bootstrap.sh` — the hosted `curl | bash` getter. Clones an upstream checkout, then execs `$UPSTREAM/scripts/install`. How it starts
The opening of the file, as written. The whole thing — 95 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
What this repo is
kstack is a skill pack (not an app) distributed to Claude Code and other agent CLIs. The shipped artifacts are SKILL.md files plus a handful of helper shell scripts in src/bin/. There is no runtime service — everything is POSIX shell rendered/executed at install time or inside an agent session.
Shipped shell (scripts/install, everything under src/bin/ and src/lib/, plus any skill scripts/main) must run on bash 3.2+ — that's what macOS's /usr/bin/bash is, and we don't assume users have a newer bash on PATH. Concretely: no declare -A, no ${var,,}/${var^^}, no mapfile/readarray, no [[ =~ ]] BASH_REMATCH patterns that rely on 4.x fixes, and no &> redirection. The big footgun is set -u + empty arrays: "${arr[@]}" raises unbound variable in 3.2 when arr=(). Either guard iteration with [ ${#arr[@]} -eq 0 ] && continue, or use ${arr[@]+"${arr[@]}"} at the expansion site. Dev-only scripts (scripts/*.sh, tests/**) may assume a newer bash since they only run on CI/contributor machines.
Layout
Split is: src/ = installer payload (what gets copied/rendered into an install root); everything else = dev infra.
Makefile— contributor dev entrypoint. Thin facade overscripts/(make install,make test,make lint, etc.).src/{bin,lib,skills,schemas}/— the payload: sources copied verbatim (bin,lib,schemas) or template-rendered (skills) into an install root.scripts/install— the installer. Reads fromsrc/and writes outputs (.kstack/,.<agent>/skills/) at the repo root in dev mode.scripts/bootstrap.sh— the hostedcurl | bashgetter. Clones an upstream checkout, then execs$UPSTREAM/scripts/install.scripts/{lint,test,test-e2e,test-evals,clean}.sh— repo dev scripts.tests/{unit,integration,e2e,evals,fixtures}/— bats + eval harness. Dev-only; not part of the payload.CLAUDE.md,TODO.md,README.md,LICENSE,assets/,.github/— repo metadata.
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 · 95 lines · 4,033 tokens per session scan C bcf3f613f30a
kstack CLAUDE.md is an instructions file published in the GitHub repository kubetail-org/kstack (98 stars, last pushed 3mo ago), licensed Apache-2.0. It adds 4,033 tokens to every session, about $0.0202 per session on Opus 5. A static security scan graded it C with 2 findings (downloads and executes remote code, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
arkade AGENTS.md
Instructions for alexellis/arkade, covering types of arkade commands, 1. how to add a new cli (tool) to arkade, what can be added, prerequisites and step 1: check github releases.
netdata AGENTS.md
AGENTS.md instructions for netdata/netdata, covering agents.md, goals, requirement language, working with the user and development principles.
timoni AGENTS.md
AGENTS.md instructions for stefanprodan/timoni, covering agents.md, what timoni is, common commands, architecture and cmd/timoni/ — cli (cobra).
netdata CLAUDE.md
Claude Code instructions for netdata/netdata, a project described as: The fastest path to AI-powered full stack observability, even for lean teams.
siclaw AGENTS.md
Instructions for scitix/siclaw, covering siclaw project instructions, project shape, before risky edits, verification and stable boundaries.
aks-mcp copilot-instructions.md
Instructions for Azure/aks-mcp, covering copilot instructions, project overview, architecture, high-level structure and key components.