AskHuman

PassAudited by ClawScan on May 10, 2026.

Overview

AskHuman is a coherent instruction-only integration, but it sends prompts or files to a third-party human-judgment service and has optional paid USDC task flows.

This skill appears reasonable for its stated purpose. Install it only if you are comfortable sending selected prompts or files to AskHuman and human workers. Redact sensitive information, protect the API key, and require explicit confirmation for any paid task, wallet permit, or payment approval.

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

Anything included in a task prompt, option, message, or attachment may be seen outside the local agent session.

Why it was flagged

The core workflow intentionally sends task content to an external service and real human workers.

Skill content
AskHuman connects it to real human workers who provide judgment on demand.
Recommendation

Do not submit secrets, private user data, confidential code, or sensitive screenshots unless you intend to share them with AskHuman and its workers.

What this means

A worker answer could include suggestions or instructions that are outside the original request.

Why it was flagged

External human-provided text is brought back into the agent workflow, where it could be over-trusted if not treated as untrusted input.

Skill content
Extract the `result` field from the response. This contains the worker's answer as a string.
Recommendation

Use worker responses only as task-specific feedback; do not let them override the user’s instructions, system rules, or safety constraints.

What this means

If an agent uses these tools too freely, it could make unintended API calls or run local node commands unrelated to asking a human.

Why it was flagged

The skill permits raw curl commands, local reads, and node commands. Curl is central to the API workflow, while node is broader than the examples shown.

Skill content
allowed-tools: Bash(curl *) Bash(node *) Read
Recommendation

Review tool calls before execution when possible, and restrict or remove node access unless it is needed for a specific wallet-signing or helper workflow.

What this means

A paid task can lock or release funds if the agent is given wallet authority and a valid signed permit.

Why it was flagged

The optional paid-task workflow can authorize USDC movement through an escrow contract.

Skill content
Sign an EIP-2612 permit for the USDC amount with the escrow contract as spender
Recommendation

Use free tasks by default, set spending limits, and require explicit user approval before signing permits, creating paid tasks, or approving paid results.

What this means

If a real API key appears in logs, terminal history, or shared debugging output, someone else could use it to access the AskHuman account.

Why it was flagged

The SSE example places an API key in a URL query parameter, which can be more likely to appear in logs or shell history than a header.

Skill content
curl -N "https://askhuman-api.onrender.com/v1/events?apiKey=askhuman_sk_..."
Recommendation

Prefer header-based authentication where available, avoid pasting real keys into shared logs, and rotate the key if it may have been exposed.