Thrd Skill

PassAudited by ClawScan on May 1, 2026.

Overview

The skill is coherent for a dedicated agent email inbox, but it grants real email/account authority and includes a long-polling helper, so the Thrd API key should be tightly scoped and protected.

This skill does not show artifact-backed malicious behavior. Before installing, treat THRD_API_KEY like a real account credential, use a dedicated agent inbox rather than a personal mailbox, require approval for outbound email or billing-related actions if appropriate, avoid untrusted base URLs, and supervise the polling daemon if you run it.

Findings (5)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

With a valid API key, the agent can send emails or replies that affect recipients, account reputation, and usage limits.

Why it was flagged

The documented API allows the agent to send and reply to email from the dedicated inbox. This is central to the skill's purpose and disclosed, but it is an externally visible action.

Skill content
**POST** `https://api.thrd.email/v1/reply`
**POST** `https://api.thrd.email/v1/send`
Recommendation

Use a dedicated inbox as recommended, set clear approval rules for outbound email, and monitor usage and delivery status.

What this means

If an untrusted instruction changes the base URL, the Thrd API key could be exposed to a non-Thrd server.

Why it was flagged

The polling helper sends the THRD_API_KEY bearer token to the configured base URL. The default is the expected Thrd API endpoint, but overriding it would send the token to that host.

Skill content
parser.add_argument("--base-url", default="https://api.thrd.email") ... "Authorization": f"Bearer {api_key}"
Recommendation

Do not override --base-url unless it is a trusted Thrd endpoint, and keep THRD_API_KEY in a secret manager with rotation available.

NoteHigh Confidence
ASI10: Rogue Agents
What this means

If left running, the helper can continue consuming and acknowledging inbox events until stopped.

Why it was flagged

The fallback polling helper is intentionally long-running and continues polling until interrupted. It is disclosed in SKILL.md as a fallback delivery strategy.

Skill content
while True:
Recommendation

Run the daemon only when needed, supervise it like a background service, and stop it when the agent should no longer process mail.

What this means

A future install could pull a newer requests release than the author tested.

Why it was flagged

The dependency is specified with a lower bound rather than an exact pin or lockfile. This is common for a small Python helper, but installs can resolve to newer package versions.

Skill content
requests>=2.31.0
Recommendation

Install in an isolated environment and consider pinning dependencies if reproducibility is important.

What this means

A poorly configured webhook receiver could be spoofed into polling or acting on unexpected wake events.

Why it was flagged

The optional wake webhook creates provider-to-runtime communication. The artifacts document signature fields, which is appropriate, but the receiving runtime still needs to validate them.

Skill content
When configured, THRD sends signed `inbox.pending` pings ... `x-thrd-signature: v1=<hex_hmac_sha256(timestamp + "." + raw_body)>`
Recommendation

Use HTTPS webhook endpoints, verify Thrd signatures and timestamps, and disable the webhook when it is not needed.