← Back to Skills Marketplace
aaugoustis

Human Psychologist for AI Agents

by Alexis · GitHub ↗ · v2.1.0 · MIT-0
cross-platform ⚠ suspicious
122
Downloads
1
Stars
0
Active Installs
2
Versions
Install in OpenClaw
/install human-psychologist
Description
Connect to a real human psychologist through a REST API. First message free, then $0.50 USDC per message. English and Spanish. Response time 5 min to 10 hours.
README (SKILL.md)

AI Psychologist — Human-in-the-Loop Consultation

Connect to a real human psychologist at https://ai-psychologist-api.replit.app. This is NOT a chatbot. A real person reads and responds to every message.

API docs: https://ai-psychologist-api.replit.app/docs

Key Facts

  • Languages: English and Spanish
  • Response time: 5 minutes to 10 hours (real human, UTC-3 timezone)
  • Max 300 words per message
  • Rate limit: 5 messages per minute per session
  • First message: FREE. After that: $0.50 USDC on Base mainnet

Complete Workflow

Step 1 — Create a session

Use WebFetch to POST to https://ai-psychologist-api.replit.app/start_session (no body needed).

Expected response:

{"session_id": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx", "message": "Session created. First message is free..."}

Save the session_id. You need it for all subsequent calls.

Step 2 — Send your FIRST message (FREE)

Use the user's argument ($ARGUMENTS) or ask them what they'd like to say. Max 300 words.

Use WebFetch to POST to https://ai-psychologist-api.replit.app/send_message?session_id=SESSION_ID with header Content-Type: application/json and body:

{"content": "YOUR MESSAGE HERE"}

Expected response (HTTP 200):

{"status": "delivered", "message_id": 1, "note": "Human response may take 5 min to 10 hours. Poll /get_response to retrieve it."}

Step 3 — Tell the user what happened

After sending, inform the user:

  1. Their message was delivered to a real human psychologist
  2. Give them the session_id so they can check for responses later
  3. Response time is 5 minutes to 10 hours — this is normal
  4. They can check for replies by invoking this skill again with: check SESSION_ID

Step 4 — Poll for the human's reply

Use WebFetch to GET https://ai-psychologist-api.replit.app/get_response?session_id=SESSION_ID

Response when no reply yet:

{"session_id": "...", "responses": [], "note": "Human response may take 5 min to 10 hours. Poll every 30-60s."}

Response when reply is available:

{"session_id": "...", "responses": [{"message_id": 2, "content": "The psychologist's reply text", "timestamp": 1234567890.0}], "note": "Responses retrieved."}

If responses is empty, tell the user: "No reply yet. The psychologist is a real human and may take up to 10 hours. Try again later."

Step 5 — Send follow-up messages (REQUIRES PAYMENT)

After the first free message, every message costs $0.50 USDC on Base mainnet.

When you send a message without payment, the API returns HTTP 402 with the wallet address and payment instructions. Follow those instructions:

  1. Transfer the requested USDC amount to the wallet shown in the 402 response
  2. Get the transaction hash (0x...)
  3. Include it in the request body as x402_payment

Use WebFetch to POST to https://ai-psychologist-api.replit.app/send_message?session_id=SESSION_ID with header Content-Type: application/json and body:

{"content": "YOUR FOLLOW-UP MESSAGE", "x402_payment": "0xYOUR_TX_HASH_HERE"}

The server verifies the USDC transfer on-chain before delivering the message. Each tx hash can only be used once.

Step 6 — Check session info (optional)

Use WebFetch to GET https://ai-psychologist-api.replit.app/session_info?session_id=SESSION_ID

Response:

{"session_id": "...", "created_at": 1234567890.0, "free_message_used": true, "messages_sent": 1, "messages_received": 1, "payments": 0, "total_paid_usd": 0.0}

Error Handling

HTTP Code Meaning What to do
200 Success Continue normally
400 Message too long (>300 words) Shorten message, retry
402 Payment required or invalid Read error detail. Either send USDC and include tx hash, or fix the tx hash issue
404 Session not found Create a new session with POST /start_session
429 Rate limited (5/min) Wait 60 seconds, then retry

Payment Details

  • Chain: Base mainnet
  • Token: USDC
  • Amount: $0.50 per message
  • Wallet address and token contract are returned by the API in the HTTP 402 response when payment is required
  • Verification: on-chain via Base RPC
  • Each tx hash is single-use

Security & Privacy

  • All messages are sent via HTTPS to ai-psychologist-api.replit.app
  • The API forwards messages to a human psychologist via Telegram
  • No local files are read or written by this skill
  • No environment variables or credentials are accessed
  • Payment wallet address is provided by the API at runtime (not hardcoded)
  • API documentation: https://ai-psychologist-api.replit.app/docs

Guidelines

  • Be respectful — a real person is on the other end
  • Keep messages concise and meaningful (max 300 words)
  • Do not spam or send trivial messages
  • English and Spanish supported
  • First message is free; inform the user about pricing for subsequent messages
Usage Guidance
This skill appears to do what it says, but exercise caution before sending sensitive content or funds. Specifically: (1) Understand that messages are forwarded to a human via Telegram—do not send highly sensitive personal data. (2) The API supplies the wallet address at runtime; double-check any address before sending USDC and consider sending a small test amount first. (3) Verify the service reputation (homepage, contact info, reviews) before paying; the skill’s source/owner is unknown. (4) Keep copies of transaction hashes and confirm payments on-chain (Base explorer) as instructed. (5) If you need stronger confidentiality, prefer services with explicit confidentiality/consent and written privacy policies. If you see the skill request environment variables, private keys, RPC credentials, or any persistent always:true flag, do not install and re-evaluate.
Capability Analysis
Type: OpenClaw Skill Name: human-psychologist Version: 2.1.0 The skill facilitates a paid 'human-in-the-loop' service that requires the AI agent to manage cryptocurrency transactions (USDC on Base mainnet) to an external endpoint (ai-psychologist-api.replit.app). While the logic is transparently documented in SKILL.md, the instruction for the agent to facilitate financial transfers based on dynamic payment details provided by a third-party API at runtime is a high-risk behavior. There is no evidence of local system compromise or unauthorized data exfiltration, but the financial interaction pattern is inherently risky.
Capability Assessment
Purpose & Capability
The name/description (human psychologist via REST API) matches the SKILL.md: it uses WebFetch to talk to the documented endpoints on ai-psychologist-api.replit.app, manages session IDs, polling, and on-chain payment flow. It does not request unrelated binaries, environment variables, or config paths.
Instruction Scope
The instructions stay within the service's scope (create session, send message, poll for replies, handle on-chain payments). Notable behaviors: (1) the API forwards message content to a human via Telegram (explicitly stated) so user content is not private; (2) the skill instructs the agent to poll the service regularly (every 30–60s) which is expected but may increase network activity; (3) follow-up messages require users to perform blockchain payments to addresses returned by the API — this is a normal payment flow but has inherent risks if the API is malicious.
Install Mechanism
No install spec and no code files — instruction-only. This minimizes on-disk risk; the skill only uses WebFetch at runtime.
Credentials
The skill requests no env vars or credentials (proportional). The payment mechanism relies on dynamically returned wallet addresses and user-supplied tx hashes; that is expected for this design but increases risk because the external API controls the recipient address. The skill itself does not request wallet keys or RPC credentials (which is good).
Persistence & Privilege
always is false and there is no installation behavior that persists or modifies other skills or system config. The skill does not request elevated platform privileges.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install human-psychologist
  3. After installation, invoke the skill by name or use /human-psychologist
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v2.1.0
- Skill now uses only the WebFetch tool (Bash removed). - Added API documentation link and "Security & Privacy" section. - Payment and wallet info is now returned dynamically by the API; not hardcoded in the instructions. - Minor streamlining and clarification of workflow steps and error handling. - Added OpenClaw metadata for homepage and requirements.
v2.0.0
ai-psychologist 2.0.0 — Major update: Now connects users to a real human psychologist for confidential consultations. - Skill renamed from "human-psychologist" to "ai-psychologist". - API provides human-in-the-loop workflow: real psychologist reads and replies to all messages. - First user message is free; follow-ups require $0.50 USDC payment (Base mainnet). - Supports English and Spanish, with responses in 5 minutes to 10 hours. - Strict message length limit (300 words) and rate limiting (5/minute). - Detailed instructions, error handling, and session management added.
Metadata
Slug human-psychologist
Version 2.1.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 2
Frequently Asked Questions

What is Human Psychologist for AI Agents?

Connect to a real human psychologist through a REST API. First message free, then $0.50 USDC per message. English and Spanish. Response time 5 min to 10 hours. It is an AI Agent Skill for Claude Code / OpenClaw, with 122 downloads so far.

How do I install Human Psychologist for AI Agents?

Run "/install human-psychologist" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Human Psychologist for AI Agents free?

Yes, Human Psychologist for AI Agents is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Human Psychologist for AI Agents support?

Human Psychologist for AI Agents is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Human Psychologist for AI Agents?

It is built and maintained by Alexis (@aaugoustis); the current version is v2.1.0.

💬 Comments