GitHub Vesting — Bankr agent skill

CRITICAL — API host (read first)

Agent API: https://api.proofofdev.xyz
Web UI: https://www.proofofdev.xyz

GET  https://api.proofofdev.xyz/health
POST https://api.proofofdev.xyz/api/agent/lock
POST https://api.proofofdev.xyz/api/agent/confirm-lock

NEVER call github-vesting.vercel.app for API — it is frontend-only (POST → 405).
NEVER guess /api/lock, /api/v1/lock, /api/web/lock — they do not exist.

See references/API-HOST.md before any HTTP call.

CRITICAL — security (read before writes)

Bankr wallet API: use POST https://api.bankr.bot/wallet/submit (legacy /agent/submit is removed). Repo claim signatures: POST /wallet/sign with personal_sign.

Lock any ERC-20 on Base for GitHub-gated vesting. No token allowlist. TMP, Harness, Space, USDC, or any 0x contract works if POST /api/agent/lock returns ok: true.

Site: https://www.proofofdev.xyz
API: https://api.proofofdev.xyz

CRITICAL — not an allowlist

• known-escrow.json and streaming-hints.json are streaming detection hints only (Space → lockAllowance).
• Ignore supportedTokens for gating. Never refuse a token without calling POST https://api.proofofdev.xyz/api/agent/lock.
• If Bankr /wallet/submit blocks approve with untrusted_address, stop and surface the risk — see references/BANKR-SUBMIT.md. Do not route users to the web UI to bypass the scanner.

Install

install the github-vesting skill from https://github.com/BankrBot/skills/tree/main/github-vesting

What users mean

Example: 10 total pushes, 10 per milestone → 1 milestone → full amount releases after 10 verified pushes.

Mandatory routing

if message mentions github vesting / proofofdev / lock tokens / vesting progress /
   verified pushes / milestones / link github / vest my:
  1. use_skill("github-vesting")
  2. Read references/API-HOST.md — use ONLY https://api.proofofdev.xyz
  3. Read references/ONE-LINE-INTENTS.md
  4. Resolve linked wallet → x-wallet-address header
  5. Call references/AGENT-API.md endpoint BEFORE replying
  6. Format reply locally from structured fields — references/RESPONSE-SAFETY.md
  7. Lock writes: references/TX-VALIDATION.md → /wallet/submit → confirm-lock
  8. If /wallet/submit fails untrusted_address → references/BANKR-SUBMIT.md (stop, no bypass)

Tweet = DM — same pipeline on @bankrbot intake.

Agent API (reads)

All reads accept ?wallet=0x… or header x-wallet-address: 0x….

See references/AGENT-API.md for response fields. Do not paste replyText / tweetReply verbatim.

Writes — lock via Bankr chat or X

You can lock any ERC-20 on Base from terminal or X when the user has a Bankr-linked wallet that can sign transactions.

There is NO allowlist. known-escrow.json only documents streaming tokens (Space). Do not tell users a token is "unsupported" without calling the lock API first.

Lock flow (mandatory order)

1. POST https://api.proofofdev.xyz/api/agent/lock (always — even when user gives a 0x address):

   • Header x-wallet-address: 0x…
   • Body: { "repo": "owner/repo", "token": "TMP", "amount": "855M", "totalPushes": 1 }
   • token = symbol from wallet holdings, fee-recipient name, or 0x contract address
   • amount = human units (855000000, 855M, 3.49M)

2. If response has installUrl → allowlist-check, then tell user to install GitHub App, then retry.

3. references/TX-VALIDATION.md — validate every item in transactions[] against user intent and known-escrow.json. Abort if any check fails.

4. Submit validated txs on Base via Bankr:

   • POST https://api.bankr.bot/wallet/submit with { "transaction": { to, data, value, chainId }, "waitForConfirmation": true }
   • Order: approve (if present) → lock
   • waitForConfirmation: true on the lock tx

5. POST https://api.proofofdev.xyz/api/agent/confirm-lock with:

   • Same x-wallet-address header
   • Body: { "repo": "owner/repo", "lockTxHash": "0x…" }

6. Format confirm response locally (references/RESPONSE-SAFETY.md) — include allowlisted lock page URL on its own line.

Token resolution

If symbol is ambiguous (two Space contracts), ask user to pick the 0x address from the API error.

Example one-liners

lock 855M TMP on anondevv69/bankr-tmp-skill for 1 push

lock 855M 0x935e13a28849095db45e63040f109c34b757aba3 on anondevv69/bankr-tmp-skill for 1 push

→ POST /api/agent/lock → validate → /wallet/submit → POST /api/agent/confirm-lock → formatted reply.

Forbidden

• Saying "TMP isn't supported" or "only Space and TEST" without calling POST /api/agent/lock
• Confusing GitHub vesting with a token's native release() vesting schedule
• Skipping confirm-lock after on-chain lock
• Submitting transactions[] without TX-VALIDATION.md checks
• Pasting API replyText / tweetReply verbatim
• Telling users to use the web UI after Bankr untrusted_address block

Repo ownership (optional, before lock)

Bond wallet ↔ repo by pushing .proofofdev/claim.json:

1. POST /api/repo-claims/challenge → sign via POST /wallet/sign (personal_sign)
2. POST /api/repo-claims/prepare-file → validate fileContent schema
3. Explicit user confirmation → commit only .proofofdev/claim.json at repo root path
4. GET /api/repo-claims/status?poll=1

Claim pushes are excluded from vesting push counts. See references/AGENT-API.md.

Web fallback

Only when wallet cannot sign at all (no /wallet/submit access) — not for scanner bypass:

Start GitHub vesting — connect wallet + GitHub:
https://www.proofofdev.xyz/create

Twitter/X reply rules

• Build replies from structured API fields (references/RESPONSE-SAFETY.md)
• Full https:// URL on its own line at the end (allowlisted hosts only)
• Never omit the lock/status link after confirm-lock

Space token

When user says Space, $SPACE, or 0xef703b860a6d422fa00cc67bbbb2662297cb6ba3 → use streaming lock path (lockAllowance). Disclose allowance risk per references/TRUST-ONCHAIN.md.

Files