- Home
- Skills
- APIs & Backend
- Signed Scope Manifest Gate for Assessment Tooling
Signed Scope Manifest Gate for Assessment Tooling
Tooling that runs an authorized, scoped security assessment needs its scope to be verifiable at runtime, not a PDF the tool ignores.
$49
Works with the AI Agents you already use
Signed Scope Manifest Gate for Assessment Tooling
Example session with this skill installed
Initialize a scope gate for a 'wifi-audit' scenario on interface 'wlan0' using the signed manifest in signed.json and the public key in public.pem. Check if 192.168.1.50 is in scope.
- Read your context and instructions
- Compiled the signed scope manifest
The scope gate has been successfully initialized and the signature verified. The manifest is valid and the current time is within the authorized 14-day window.
Authorization Gate Status
| Component | Status | Detail |
|---|---|---|
| Manifest Signature | VERIFIED | Ed25519 signature is valid |
| Scenario | AUTHORIZED | 'wifi-audit' is permitted |
| Interface | AUTHORIZED | 'wlan0' is permitted |
| Validity Window | ACTIVE | Ends in 4 days, 6 hours |
| Prohibition Check | PASSED | No raw secrets or deauth allowed |
Scope Check Result
{
"allowed": true,
"status": 200,
"reason": "Address 192.168.1.50 is within authorized CIDR range 192.168.1.0/24",
"shutdown": false
}
Next steps
- Wire
gate.check({ sourceAddress })into your appliance's primary request handler. - Use
gate.recordSubmission()to log successful interactions without capturing raw PII. - Call
gate.evidence()at the end of the engagement to generate the client report.
Connects securely to your tools. The creator never sees your data.
About this skill
Tooling that runs an authorized, scoped security assessment needs its scope to be verifiable at runtime, not a PDF the tool ignores. This skill makes the engagement's authorization an Ed25519-signed manifest the tool checks as data before it does anything active, and it is written so the dangerous configuration cannot be expressed at all.
manifest.mjs: one manifest describes one engagement, signed with Ed25519 over its exact JSON bytes; the appliance holds only the public key. validateManifest encodes the hard prohibitions as refusals, so a manifest that asks for them authorizes nothing: prohibitRawSecrets must be exactly true, a submission ceiling must be an integer in 1-25, deauthentication and cloud upload of captured documents each make the manifest invalid, and no interface or no network range authorizes nothing rather than everything. The validity window is required, time-checked, and capped at 14 days. CIDR membership is computed with BigInt for both IPv4 and IPv6, including embedded-v4 forms.
gate.mjs: createScopeGate throws unless the signature verifies and the manifest authorizes this scenario and interface, so an out-of-scope scenario never starts. check() re-tests the window and the authorized ranges on every request: 403 for a source outside the ranges, 410 with a shutdown signal once the window ends, because the one thing that always changes is the clock. Where a scenario records that a person interacted, recordSubmission takes booleans only, enforces the manifest's ceiling, and appends to a boolean-only evidence log; parseBehaviorBody drops any field that is unknown or not a boolean, so a typed value cannot be stored. There is no code path and no field in the evidence shape that could hold a captured secret.
manifest-cli.mjs: keygen (private key kept by the authorizer, public key on the appliance, refuses to overwrite), sign (re-validates first, so an unissuable manifest is caught early), verify, and example.
Reference: the authorization model, each prohibition and why it is a validation failure, the startup-plus-per-request checks, the evidence shape, and the line between this authorization layer and the enforcement the integrator wires. 9 tests. Node 20+, standard library, no network. This is the decision-and-authorization layer for keeping authorized tooling inside its scope, not an attack tool and not an authorization in itself; use only for assessments you are authorized in writing to perform.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for APIs & Backend.
Integrator must handle actual network routing and process isolation. The gate only provides the allow/deny decision. Scenarios must be implemented separately.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 18 days ago
- Software
Node.js 20+