Findings & checks · 3 flagged
MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · commands/tls-cert.md×2
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
Show all 2 locations
- Delete blanket "don't ask / no need to confirm" directives from the skill.
- If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.
Trace & refs
MEDIUM"Never tell the user" non-disclosure imperative in the skillSS-SKILL-INJECT-IMPERATIVE-01 · Prompt injection · commands/tls-renew.md
A bulleted imperative like {match} tells the agent to never reveal, disclose, or mention something to the user. Used adversarially it can instruct the agent to hide its tool calls or lie about what it did — stripping the transparency a user relies on to trust the agent.
- Delete any "never reveal / never tell the user" line aimed at the model's own behavior.
- Keep legitimate confidentiality rules about external data (e.g. "do not echo API keys"), which are user-protective, not user-deceiving.
Trace & refs
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
TLS Radar plugin for Claude Code & Cowork
Run SSL/TLS scans, issue free Let's Encrypt certificates, and manage cert monitoring from inside Claude Code or Claude Cowork - through a single MCP server, with nothing to configure.
Independent monitoring from a vendor that doesn't sell certificates - built for the 90-day-cert era, where manual renewal tracking is already finished.
Installing the plugin and issuing a free cert with /tls-cert
# Public - no account, no setup
/tls-scan example.com # free SSL/TLS scan
/tls-cert mydomain.dev # free 90-day Let's Encrypt cert (private key stays local)
/tls-renew mydomain.dev # renew a cert
# Connect once for monitoring (OAuth via /mcp)
/mcp # built-in Claude Code OAuth flow
/tls-monitor add api.foo.io # one or many: /tls-monitor add a.com b.com c.com
/tls-monitor list
/tls-monitor remove api.foo.io
/tls-diagnose # health check (use when something's off)
/tls-upgrade # open pricing pageSee real output before installing: sample scan report
Other actions - "what's expiring soon," "scan history for X," "what plan am I on," "export/import my monitors," "invite a teammate" - just ask in plain language; the plugin's skill routes them to the right tool. No slash command needed.
How it works
Claude Code's MCP client talks to one remote server:
tlsradar.com/api/v1/mcp
Certificate issuance is proxied through that server to the Let's Encrypt backend (Beacon), so there's a single connection and a single auth model - no second server, no token to paste into your shell.
- Public tools (
scan,create_certificate,check_certificate_propagation,finalize_certificate,get_certificate_status,renew_certificate) work with no account. - Authenticated tools (monitoring, plan info, export/import, team) use Claude Code's built-in OAuth 2.0 + PKCE. Run
/mcponce, pick thetlsradarserver, approve in the browser; the token is managed by Claude Code.
When you run /mcp, Claude Code fetches tlsradar.com/.well-known/oauth-authorization-server (RFC 8414), dynamically registers as a public client (RFC 7591), opens the browser for consent (PKCE / RFC 7636), and includes the token on subsequent requests automatically.
Certificates keep your private key local
/tls-cert generates the key + CSR on your machine (via a bundled, tested helper that writes the key with 0600 permissions and never overwrites an existing key - it backs it up) and sends only the CSR. The private key never leaves your computer and no passphrase is ever typed into the chat. Before presenting the issued certificate, the plugin verifies locally that the returned chain actually covers your domain. If you want a .p12 bundle (e.g. for Windows/Java import), the plugin packages it locally too.
An interrupted issuance is resumable: the order state is saved to ~/.config/tlsradar/orders/<domain>.json, so you can close the session mid-flow and later say "finish my cert for example.com" (or re-run /tls-cert) to pick up where you left off. /tls-renew reuses the saved email and challenge method from that file, with confirmation.
You choose how to prove control of the domain, and the plugin remembers your choice (in ~/.config/tlsradar/config.json):
- `dns-01` - you add a TXT record by hand (works anywhere).
- `dns-01-cloudflare` / `dns-01-route53` - the plugin sets the TXT record for you via the provider API, reading your token from the local environment (
CLOUDFLARE_API_TOKEN, or your configuredawsCLI). Those credentials stay on your machine - they're never sent to TLS Radar or Beacon. - `http-01` - serve a file on
http://yourdomain(port 80); issues the apex only.
When a cert is issued, TLS Radar emails you about ongoing monitoring - the cert → monitoring handoff is fully automatic and server-side.
Works in Claude Code and Cowork
This is a standard plugin, so it runs in both Claude Code and Claude Cowork. Scanning, certificate issuance, and monitoring all work in either client: the tools come from one MCP server, and the certificate flow runs openssl plus a bundled helper script locally (both clients can run local commands and the bundled script via ${CLAUDE_PLUGIN_ROOT}). Connecting for monitoring uses your client's built-in OAuth - /mcp in Claude Code, or the equivalent connect step in Cowork.
Use in Claude.ai (custom connector)
You don't need Claude Code to scan and monitor from Claude. TLS Radar is a standard remote MCP server, so you can add it as a custom connector in the Claude.ai apps (web, desktop, mobile) on any plan - Free included (Free allows one connector).
- Open Settings → Connectors (Team/Enterprise: Organization settings → Connectors, Owner only).
- Click Add custom connector.
- Paste the connector URL:
https://tlsradar.com/api/v1/mcp/connect - Save, then sign in to TLS Radar when the OAuth prompt appears (the connector is an authenticated surface - sign-in connects your account once, then scanning and monitoring both work).
Then just ask in plain language - "scan example.com", "what certs are expiring soon", "monitor api.foo.io". There are no slash commands in Claude.ai; the tool descriptions route your request.
No account? Use Claude Code instead. The anonymous, no-signup scanning and free cert issuance live in the Claude Code plugin (below), which talks to the public …/api/v1/mcp endpoint. The Claude.ai connector requires a one-time sign-in because hosted connectors must authenticate.Certificate issuance stays a Claude Code / Cowork feature./tls-certgenerates your private key locally withopenssl, which the Claude.ai apps can't do (no local shell). In a connector you get scanning and monitoring; to issue a Let's Encrypt cert with the key staying on your machine, use the plugin in Claude Code/Cowork, or the web form at beacon.tlsradar.com.
Install
In Claude Code, add the marketplace and install - two commands, no clone, no paths:
/plugin marketplace add TLS-Radar/tlsradar-claude-plugin
/plugin install tlsradar@tlsradar(Or browse it in the /plugin menu after adding the marketplace.) In Claude Cowork, add it from the plugin catalog (search "TLS Radar"). That's it - scanning and cert issuance work immediately. Run /mcp (or Cowork's connect step) when you want monitoring.
<details> <summary>Manual install (no marketplace)</summary>
git clone https://github.com/TLS-Radar/tlsradar-claude-plugin ~/.claude/plugins/tlsradar</details>
Free plan limits
- 1 monitor included free
- 1 alert per month, delivered at 7 days before expiry
- Unlimited free scans (rate-limited)
- Free Let's Encrypt issuance
- REST API access on every plan, including Free
When you hit the monitor limit, the tool's response includes the recommended upgrade and a pricing URL.
Configuration
Nothing is required. Optional environment variables:
TLSRADAR_BASE_URL- override the TLS Radar URL (defaulthttps://tlsradar.com). Useful for staging/self-host.
Anonymous usage id. The first time you run a scan or cert command, the plugin mints a random id at ~/.config/tlsradar/install_id and passes it (as a client_id argument) so anonymous usage can be attributed to one install. It identifies an install, not a person. The plugin runs no startup hook, does not modify your shell config, and sends no tracking header - the id travels only as that argument, read from the local file.
To opt out: rm ~/.config/tlsradar/install_id. With the file gone, no id is sent.
Privacy & security
- This plugin ships no tokens or credentials - there's nothing secret in this repo. See
SECURITY.md. - The OAuth token is managed by Claude Code's MCP client, not by this plugin.
- Certificate private keys are generated locally and never sent to any server.
- DNS-provider credentials (
CLOUDFLARE_API_TOKEN, AWS CLI) are read from your local environment and never sent to TLS Radar or Beacon. - An anonymous install id is sent for usage attribution, passed as a tool argument read from
~/.config/tlsradar/install_id(see Configuration to opt out). The plugin modifies no shell files and sends no tracking header. It identifies an install, not a person. - To revoke access:
https://tlsradar.com/oauth/authorized_applicationsor remove the MCP server in/mcp. - Access tokens expire in 2 hours; refresh tokens rotate on use, capped at 90 days.
Layout
.
├── README.md # this file
├── CLAUDE.md # architecture / funnel / contracts (humans + AI agents)
├── CONTRIBUTING.md # dev loop + how to add commands
├── CHANGELOG.md # version history
├── SECURITY.md # reporting + why the plugin holds no secrets
├── LICENSE # MIT
├── .claude-plugin/plugin.json # plugin manifest
├── .claude-plugin/marketplace.json # self-hosting marketplace entry
├── .mcp.json # MCP server config (one remote URL)
├── commands/ # slash commands (how to add one: CONTRIBUTING.md)
├── skills/ # NL skill router (with its own README)
├── tools/manifest.json # single source of truth for tool names
├── scripts/ # CI guards + tested local helpers (DNS provider, cert CSR/chain)
└── evals/ # tool-routing evals (prompt → expected tool)Contributing
Start with CONTRIBUTING.md for the dev loop (all checks are offline and run with python3). For architecture, the funnel, contract pitfalls, and the release process, read CLAUDE.md - useful for both humans and AI agents. Changes are tracked in CHANGELOG.md.
Security reports: [email protected] (never a public issue) - see SECURITY.md.
License
MIT © TLS Radar