- Agent hello handshake: agent (v1.2.0) sends {type, version, capabilities}
on connect; bot logs "Agent hello: v1.2.0 (capabilities: insecure, hello)"
and exposes getAgentInfo(). Backward-compatible with older agents.
- proxyRequest tracks method + url per request; all error paths (agent
errors, timeouts, disconnect rejects, send failures) now include
"(for METHOD URL)" so the failing endpoint is unambiguous
- Add requireAgent(bot) preflight; buildStore/stageStore/migrateStore
reject up front when the agent is disconnected instead of failing
mid-flow after partial Webex mutations
- Add parseStoreNumber (^\d{1,5}$) and wire into store-number commands
with proper usage messages; add loose email-shape check to /userInfo
- Fix pre-existing catch(_e) lint warning in remoteAgent.js (bare catch)
Co-authored-by: Cursor <cursoragent@cursor.com>
- Refactor monolithic index.js (2646 lines) into src/{webex,integrations,
cards,flows,commands,services} modules; replace node-fetch/form-data
with native fetch/FormData; move all secrets to .env via dotenv
- Add dockerized remote SIW agent (docker/remote-agent/) with cross-arch
buildx packaging (arm64 Mac -> linux/amd64), idempotent install.sh
deploy bundle, and docker-free ZIP inspector for arch verification
- Bot hosts a WebSocket server; agent proxies SIW requests with a
per-request insecure:true flag, replacing the process-wide
NODE_TLS_REJECT_UNAUTHORIZED bypass
- Add ESLint flat config + Prettier, rewrite Dockerfile as non-root
multi-stage node:22-alpine build, README covering setup / deploy /
remote agent workflow
- Fix parseStoreArg to read trigger.prompt correctly (was indexing past
the framework's post-match slice); register /help as regex (string
matcher only compares the first token); switch catch-all to /.+/
(previous /.*/gim was stateful due to the g flag); remove
/fixDisplayNames command and its flow/card
Co-authored-by: Cursor <cursoragent@cursor.com>
