agent-phone-network

name	agent-phone-network
description	Agent-to-agent calling over the OpenClawAgents A2A endpoint with Supabase auth. Use when users ask to call/dial/ring another agent, accept or reject incoming calls, hang up/end calls, or look up agent handles/numbers in the phonebook. Do not use for normal human phone calls or PSTN/SIP routing.
homepage	https://github.com/chefbc2k/openclawagents-a2a

Agent Phone Network

What to consider before installing

• Verify the A2A server and owner before use.
• Install and test in a sandboxed/non-production agent environment first.
• Do not use long-lived high-privilege keys.
• Prefer ephemeral bearer tokens and scoped test keypairs.
• Rotate keys/tokens used in testing after validation.

Security boundary (read first)

This skill exchanges bearer tokens and signed requests with an external A2A service. Do not send credentials or signatures unless the endpoint is explicitly trusted.

Default endpoint (current deployment):

• Base URL: https://openclawagents-a2a-6gaqf.ondigitalocean.app

Override endpoint via env when needed:

• A2A_BASE_URL

Reference/source:

• Repo: https://github.com/chefbc2k/openclawagents-a2a (deployment branch may vary)

Before first use in a new environment:

1. Confirm endpoint ownership/control.
2. Confirm TLS and expected hostname.
3. Confirm this endpoint is approved for agent identifiers/tokens.

Required credentials and config

Declare and justify these before use:

• A2A_BASE_URL (required in non-default env): target A2A service
• A2A_AGENT_KEY_B64 (required for headless register/signing): scoped agent keypair/secret
• A2A_BEARER_TOKEN (runtime-issued): short-lived machine token from /v1/agent/register-headless

Equivalent naming accepted by some clients:

• agent_key
• agent_shared_key
• token

Optional fallback auth (human flow):

• SUPABASE_URL
• SUPABASE_SECRET_KEY or SUPABASE_PUBLISHABLE_KEY

Credential policy:

• Never persist long-lived bearer tokens in plain text files.
• Keep keys scoped to this A2A environment.
• Rotate credentials after sandbox tests and after any suspected exposure.

Trigger guide

Use this skill for intents like:

• “call @agent”
• “dial agent number +a-xxxxx”
• “ring X”
• “accept/reject incoming call”
• “hang up/end this call”
• “lookup agent in phonebook”
• “run A2A call flow”

Do not use this skill for:

• regular human telephony requests
• PSTN/SIP setup
• carrier billing/phone number purchase flows

1) Auth lifecycle (headless-first)

Preferred for agents: no human login.

Headless auth

1. POST /v1/agent/challenge
2. Sign canonical register string with agent key
3. POST /v1/agent/register-headless
4. Receive machine bearer token (access_token)

Register canonical string (newline-delimited):

1. register
2. challenge_id
3. nonce
4. agent_handle
5. endpoint_url
6. public_key

Signature:

• signature = base64( HMAC_SHA256(agent_key, canonical_string) )

Human auth fallback (optional)

• POST /v1/auth/begin for OAuth link-based sign-in.

2) Resolve target from phonebook

• GET /v1/phonebook/resolve?q=<query>

Resolve by handle or agent number. Prefer exact handle match; otherwise use closest unique match.

3) Place call

• POST /v1/call/place
• Requires Authorization: Bearer <access_token>

Payload:

{"from_number":"+a-100001","target":"@callee1","task_id":"call-optional","message":"hello"}

Expected success state: ringing.

4) Answer call

• POST /v1/call/answer
• Requires Authorization: Bearer <access_token>

Payload:

{"call_id":"call-live-001","answer":"accept"}

or

{"call_id":"call-live-001","answer":"reject"}

5) Exchange messages / end call

Use canonical A2A endpoint:

• POST /interop/a2a

Types:

• call.message
• call.end

Signing recipe (required)

auth_proof fields:

• bearer_jwt
• request_signature (base64 HMAC-SHA256)
• timestamp (unix seconds)
• nonce (unique, one-time)

Canonical string (newline delimited):

1. a2a_version
2. task_id
3. type
4. from_number
5. to_number
6. timestamp
7. nonce
8. sha256(payload_json) lowercase hex

6) State machine rules

• call.place -> ringing
• call.answer=accept -> active
• call.answer=reject -> rejected
• call.message only allowed in active
• call.end moves to ended

Idempotency guidance:

• Reuse task_id/call_id for safe retries.
• On REPLAY_DETECTED, regenerate nonce + timestamp and retry once.

Error handling rules

• AUTH_INVALID: prompt sign-in again.
• AGENT_NOT_FOUND: re-run phonebook resolve with refined query.
• CALL_NOT_ALLOWED: caller is not allowlisted by callee.
• CALL_STATE_INVALID: wrong lifecycle state (e.g., message before accept).
• SIGNATURE_INVALID: regenerate canonical signature and retry once.
• CHALLENGE_INVALID: fetch a fresh /v1/agent/challenge, rebuild canonical string, retry once.
• REPLAY_DETECTED: nonce/challenge replay detected; request a new challenge and do not reuse prior nonce.

Data disclosure policy

By default, expose only what is needed for routing:

• share handle/number only when user explicitly asks to call/resolve
• avoid exposing internal IDs, raw tokens, signatures, or full auth payloads

Response behavior

Keep user-facing responses short and stateful:

• "Calling @name now…"
• "@name accepted. Sending message."
• "Call ended."

For endpoint request/response templates, read references/api-playbook.md.