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 skills add arcjet/arcjet-js --skill integrate-arcjet-guard-claude-managed-agentsgit clone --depth 1 https://github.com/arcjet/arcjet-jsWrote 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/skills/arcjet/arcjet-js/integrate-arcjet-guard-claude-managed-agents)<a href="https://agentmods.dev/skills/arcjet/arcjet-js/integrate-arcjet-guard-claude-managed-agents"><img src="https://agentmods.dev/badge/skills/arcjet/arcjet-js/integrate-arcjet-guard-claude-managed-agents/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/arcjet/arcjet-js/integrate-arcjet-guard-claude-managed-agents"><img src="https://agentmods.dev/badge/skills/arcjet/arcjet-js/integrate-arcjet-guard-claude-managed-agents.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 5.1 | $0.00102 | $0.02463 |
| Opus 5.5 | $0.00041 | $0.00985 |
| Sonnet 5.5 | $0.00020 | $0.00493 |
| Haiku 4.5 | $0.00010 | $0.00246 |
Grade A, and why
integrate-arcjet-guard-claude-managed-agents 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 7d 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 — 240 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Integrate Arcjet Guard into Claude Managed Agents
@arcjet/guard's Claude Managed Agents v0 namespace wraps the agent's
existing Arcjet client. It never talks to the Arcjet API itself. Anthropic
runs the hosted tool loop. There is no PreToolUse.
This is not @arcjet/guard/claude-agent-sdk/v0. Do not reuse that
adapter, its hooks, or guardTool.
- Inbound text (
user.message/initial_events) →guardEvents()beforesessions.events.send. DENY does not send. There is noguardInbound. - A custom tool you execute (
agent.custom_tool_use) →guardCustomTool(). DENY does not run the tool; senduser.custom_tool_resultwithis_error: trueand error text (that field exists on the real events API). Self-hostedEnvironmentWorker/betaTool({ run })uses the same gate. The CLI worker cannot register custom tools. - Correlation →
claudeManagedAgentsContext()is caller-owned only. It never mints. It never treats Anthropic session/event ids as if we created them. NevertraceId.
What cannot be gated
Default permission_policy: always_allow cannot be gated. Anthropic
executes bash/read/write in the cloud sandbox before your process sees an
event. web_search / web_fetch always run on Anthropic. MCP: Anthropic
is the client; customer-side Guard is on custom tools and MCP servers
you host. user.tool_confirmation is HITL (always_ask), not a
policy gate — do not add a confirmation helper as the happy path.
Same trap as Cloudflare Think needsApproval.
protect() stays fail-open. These helpers are fail-closed.
Docs:
https://docs.arcjet.com/guards/claude-managed-agents/
(shared JS+Python page). Do not edit /guards/claude-agent-sdk/ or
/guards/claude-agent-sdk-py/.
Questions to ask the human first
Ask only what you cannot infer from the code; suggest defaults.
- Which custom tools does the app execute on
agent.custom_tool_use? Those getguardCustomTool. Built-ins underalways_allowcannot. - What limits? (e.g. "10 lookups/min per order" →
tokenBucket.) - Who is the user for metadata — an opaque user/tenant ID (never PII)? Default: none. Session/event ids from Anthropic are not the correlation id.
- Is an Arcjet outage unacceptable? Every helper defaults to
onGuardError: "deny", which also decides a call whoserulesormetadatacallback threw or returned a value Guard cannot use: Guard is still called without that value, and a DENY stands. Ask explicitly about inboundguardEvents: failing closed there means the turn is not sent.
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.
- 7d ago Changed · +6 lines 61d44f6937b0
- 21d ago Changed · +5 lines 38b2c350fb21
- 29d ago Changed · +2 lines cb7ea13e3cba
- 1mo ago Changed · +3 lines 5ae1911b97ad
- 1mo ago Changed 3e1d9f5fbd3a
- 1mo ago Changed · +1 lines d20c545fad8a
- 1mo ago Changed b547b2468932
- 1mo ago Changed · +6 lines b62dfc106909
- 2mo ago First seen · 217 lines · 102 tokens per session scan A c7301c097f38
integrate-arcjet-guard-claude-managed-agents is a skill published in the GitHub repository arcjet/arcjet-js (698 stars, last pushed today), licensed Apache-2.0. It adds 102 tokens to every session and 2,463 once invoked, about $0.0004 per session on Opus 5.5. 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 2026-09-05.
Other skills, from other repositories
adr-skill
Create and maintain Architecture Decision Records (ADRs) optimized for agentic coding workflows. Use when you need to propose, write, update, accept/reject, deprecate, or supersede an ADR; bootstrap an adr folder and index; consult existing ADRs before implementing changes; or enforce ADR conventions. This skill uses…
clinical-trial-protocol-skill
Generate clinical trial protocols for medical devices or drugs. This skill should be used when users say "Create a clinical trial protocol", "Generate protocol for [device/drug]", "Help me design a clinical study", "Research similar trials for [intervention]", or when developing FDA submission documentation for…
claude-d3js-skill
This skill provides guidance for creating sophisticated, interactive data visualisations using d3.js.
prior-auth-review-skill
Automate payer review of prior authorization (PA) requests. This skill should be used when users say "Review this PA request", "Process prior authorization for [procedure]", "Assess medical necessity", "Generate PA decision", or when processing clinical documentation for coverage policy validation and authorization…
linear-claude-skill
Manage Linear issues, projects, and teams.
playwright-skill
IMPORTANT - Path Resolution: This skill can be installed in different locations (plugin system, manual installation, global, or project-specific). Before executing any commands, determine the skill directory based on where you loaded this SKILL.md file, and use that path in all commands below.