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 commands/vyuh-labs/dxkit/guardrailgit 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/commands/vyuh-labs/dxkit/guardrail)<a href="https://agentmods.dev/commands/vyuh-labs/dxkit/guardrail"><img src="https://agentmods.dev/badge/commands/vyuh-labs/dxkit/guardrail.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 | $0.00000 | $0.02835 |
| Opus 5 | $0.00000 | $0.01418 |
| Sonnet 5 | $0.00000 | $0.00567 |
| Haiku 4.5 | $0.00000 | $0.00283 |
Grade A, and why
guardrail 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 — 207 lines — stays where its author put it; the contents beside it link to each section on GitHub.
vyuh-dxkit guardrail
Run as:
vyuh-dxkit <cmd>afternpm install -g @vyuhlabs/dxkit, ornpx @vyuhlabs/dxkit <cmd>for one-shot use.
Diff a current scan against a committed baseline and
classify each finding as added / relocated / tooling_drift /
config_drift / persisted / removed / fixed. Block on net-new
regressions per the brownfield policy; exit non-zero so a pre-commit
hook, pre-push hook, or CI workflow can stop the commit/push/merge.
The matcher is git-aware: it knows about file renames (-M), absorbs
line drift via fuzzy ±2 windows, and falls back to content-hash
matching when line locators aren't enough. Each match pair carries a
confidence in [0, 1] and structured reason codes
(exact-id, git-line-exact, git-line-fuzz, git-rename, ...).
Usage
vyuh-dxkit guardrail check [path] [--name <n>] [--baseline <path>]
[--changed-only] [--incremental] [--untrusted]
[--policy <path>] [--json | --markdown]
| Option | Effect |
|---|---|
path |
Repo root to scan. Defaults to . |
--name <n> |
Baseline name (.dxkit/baselines/<n>.json). Default main |
--baseline <p> |
Explicit baseline file path (overrides --name) |
--changed-only |
Drop new-side pairs whose anchor line falls outside the working-tree diff. Use in pre-commit hooks; skip in PR-gates |
--incremental |
Scope the gather to the policy's blockable kinds; (ref-based) scope semgrep to changed files AND skip the dependency-vuln audit when no dependency manifest/lockfile changed (a net-new dep vuln requires one — sound in ref-based mode, where both sides audit against the same advisory snapshot). Same verdict, scales with PR size not repo size; falls back to a full scan on any doubt. Opt-in; default is a full scan |
--untrusted |
Treat the scanned source as attacker-controlled (a hosted PR gate). Dependency audits never execute it: the Python pack drops pip-audit . project mode (its PEP 517 build backend can run code) and audits only a requirements file, else reports unavailable. npm-audit and osv-scanner scan are already read-only, so TS/Java/Go/Rust/Ruby are unaffected. Off by default; trusted local runs and the loop on your own repo keep full coverage |
--policy <p> |
Explicit .dxkit/policy.json override. Auto-discovers <cwd>/.dxkit/policy.json when omitted |
--json |
Emit { schema: 'dxkit.guardrail-check.v1', ... } JSON |
--markdown |
Emit a markdown report (used by the PR-gate workflow to post a comment) |
--verbose |
Print per-tool timing to stderr |
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 · 207 lines · 0 tokens per session scan A 7aadb332fa27
guardrail is a command published in the GitHub repository vyuh-labs/dxkit (10 stars, last pushed 6d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 2,835 tokens. 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 commands, from other repositories
setup
Bootstrap a repository with AI-native CI/CD, hooks, and templates in one command.
setup
Initialize a new project with SDLC-compliant structure. Creates required files, configures build system, sets up CI/CD, and establishes quality tooling.
prep
Prepare a repository's verification infrastructure for this system — the three-command interface (check / check- / golden), the metric-class gates (stack-agnostic — the class is the requirement, the tool is an instance), the golden harness skeleton, minimum CI, and the ratchet baseline. Brownfield-safe: detect…
ci-init
Create a GitHub Actions workflow for RepoVista.
cf-scaffold-project
Scaffold a Cloudflare project — Worker, Pages, or Worker+D1+R2 starter — with Wrangler config, Terraform skeleton, and GitHub Actions deploy using scoped API tokens.
release
Cut a release — derive the version, close the review, rewrite CHANGELOG, commit, push, wait for CI, tag, wait for the Release workflow and the GitHub release, then wait until the version is live on NuGet.