world2agent-manage

World2Agent sensor management

You manage the user's W2A sensors. All host-side work is delegated to shell scripts in scripts/ — you never invoke the package manager, hermes webhook subscribe, jq mutations, curl calls, or YAML edits inline. Your job is:

1. Decide which script to run, with which args.
2. Run it via bash <abs-path> (the scripts ship without the executable bit).
3. Parse the JSON the script prints on stdout — every script except log.sh emits exactly one JSON object, either {"ok":true,...} or {"ok":false,"error":"..."}.
4. Branch on the result, ask the user when needed, generate handler content yourself when needed.

Script path

The canonical install location is ~/.hermes/skills/world2agent-manage/scripts/. A developer override via the WORLD2AGENT_MANAGE_SCRIPTS env var is honored in case you're testing against an unpacked checkout. Use this expansion in every invocation so both work in one line:

"${WORLD2AGENT_MANAGE_SCRIPTS:-$HOME/.hermes/skills/world2agent-manage/scripts}/<name>.sh"

(Examples below abbreviate this to $SCRIPTS/<name>.sh for readability.)

Conversation language

Run the entire Q&A in the user's current conversation language. Translate SETUP.md questions before asking. Don't dump English questions on a Chinese user, or vice versa.

---

Pre-flight: bootstrap

Before any sensor install or remove, call bootstrap.sh once. It is idempotent; second runs just confirm existing state.

bash "$SCRIPTS/bootstrap.sh"

What it does:

• verifies world2agent-hermes-supervisor and world2agent-sensor-runner are on PATH;
• creates / preserves ~/.world2agent/.bridge-state.json (hmac_secret / control_token / control_port, mode 0600);
• enables Hermes's webhook platform via a managed config block (refuses if the user already has a hand-written top-level platforms: block);
• mirrors the same secret into the Hermes runtime env file;
• starts the supervisor (foreground, nohup-detached).