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/vyuh-labs/dxkit/claude-mdgit clone --depth 1 https://github.com/vyuh-labs/dxkitWrote 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/vyuh-labs/dxkit/claude-md)<a href="https://agentmods.dev/instructions/vyuh-labs/dxkit/claude-md"><img src="https://agentmods.dev/badge/instructions/vyuh-labs/dxkit/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 | $0.24268 | $0.24268 |
| Opus 5 | $0.12134 | $0.12134 |
| Sonnet 5 | $0.04854 | $0.04854 |
| Haiku 4.5 | $0.02427 | $0.02427 |
Grade A, and why
dxkit CLAUDE.md scanned grade A 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
CLOSED on mismatch. The arch-check bans a raw curl/wget-with-URL in Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
content-scan inside `run()` / `execSync()` outside the 4-file How it starts
The opening of the file, as written. The whole thing — 1,660 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md — DXKit Development Rules
Architecture Rules
1. Tool invocation goes through the registry
Every external tool (cloc, gitleaks, semgrep, graphify, jscpd, ruff, etc.) MUST be:
- Defined in
src/analyzers/tools/tool-registry.ts(TOOL_DEFS) - Detected via
findTool(TOOL_DEFS.xxx, cwd)— never hardcode binary paths - Installed via
vyuh-dxkit tools install— never ad-hoc npx/pip calls
Builtins (grep, find, wc, git, node) are exempt — they're always available.
Executed downloads are checksum-verified (4.0.3 / T2.1). Any network
artifact an install command fetches goes through
dxkit_fetch <url> <sha256> <dest> — defined ONCE in
src/analyzers/tools/install-exec.ts (the single install executor prepends
it), with the pinned hash declared in the registry beside the URL. It fails
CLOSED on mismatch. The arch-check bans a raw curl/wget-with-URL in
tool-registry.ts and an Invoke-WebRequest without a Get-FileHash check
(annotate // unverified-download-ok for a justified exception);
test/tool-registry-version-pins.test.ts pins that every dxkit_fetch call
carries a 64-hex sha and that no unverified fetch exists in any install
command. Bumping a tool version means re-pinning its hash (prefer the
publisher's published checksum file where one exists — gitleaks, codeql).
CI tool discoverability is registry-derived, never a hardcoded PATH. The
places a tool binary can live are declared once — getSystemPaths() plus each
tool's probePaths in the registry — and findTool probes them. The CI PATH
export (exportToolPathsToGithubEnv, run at the end of tools install) reads
the SAME sources and writes them to $GITHUB_PATH, so the per-language dep audit
finds its native scanner (osv-scanner / pip-audit / govulncheck / cargo-audit)
in a workflow step instead of silently falling back to a wrong-artifact scanner.
A new language pack that installs a scanner to a new directory declares it in the
tool's probePaths (already required for detection) and is thereby covered in CI
automatically — do NOT hardcode a per-ecosystem bin dir in a workflow template.
test/tool-paths-ci.test.ts pins that every tool probePath appears in the
export.
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 · 1,660 lines · 24,268 tokens per session scan A ae117133b1a5
dxkit CLAUDE.md is an instructions file published in the GitHub repository vyuh-labs/dxkit (10 stars, last pushed 7d ago), licensed MIT. It adds 24,268 tokens to every session, about $0.1213 per session on Opus 5. A static security scan graded it A with 2 findings (makes network calls, runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
habit-hooks CLAUDE.md
Instructions for habit-hooks/habit-hooks, covering habit-hooks notes, architecture, gotchas, a tool that resolves symlinks now hard-fails its sensor and jsdoc nodes are not multilinecommenttrivia in ts-morph.
jrag CLAUDE.md
Claude Code instructions for HumanBean17/jrag, covering jrag, python environment, tests, docs and shipped artifacts.
agent-feed AGENTS.md
Instructions for fqmyysjjd/agent-feed, covering agent-feed ai development instructions, startup policy, rule priority, responsibility boundary and mandatory gates.
agent-feed CLAUDE.md
Instructions for fqmyysjjd/agent-feed, a project described as: Turn AI coding chats into a source-controlled workflow: AGENTS.md, rules, skills, verification gates, and adapters for Codex, Claude Code, and Cursor.
claude-agentic-coding-playbook CLAUDE.md
Instructions for john-wilmes/claude-agentic-coding-playbook, covering agentic coding playbook, quality gates, architecture, dependencies and build and project conventions.
plan-forge dapr.instructions.md
Dapr patterns for .NET — building blocks, component config, sidecar architecture, multi-tenant isolation, workflows, state management, secrets.