Hinata Wiki

Appearance

Bot reply should demonstrate vault-connectivity (receipt pattern) ​

Gap ​

The 5 Commander bots (bulma, ging, zuko, brook, sanji) now reply in their Commander's voice via hinata-bot-poller.py (Anthropic API + context.md as system prompt). But the reply itself doesn't prove vault-connection — a generic "yes I'm connected" doesn't demonstrate the bot has actually read its Commander's context.

On 2026-05-26 Michael messaged three bots with "are you connected to hinata ?" / "Connect". Each bot replied, but he had no visible way to verify the reply was anchored in vault state.

What "demonstrating connectivity" looks like ​

Each Commander-bot reply should optionally surface a vault receipt — a 1-line footer (or system-detectable marker) showing:

  • Which context.md was loaded (zuko-career/context.md)
  • Current focus for that Commander (1-line summary)
  • Last-touched timestamp on the context (so freshness is visible)

Example reply with receipt:

"Yes — Zuko here, loaded from federation/colonel_levi-ackerman-forge_zuko-career_context.md (last touched 2026-05-25). Current focus: dbt cert resit + Senior DS / Quant Analyst job search at £80k+. Ask anything in scope."

Implementation options ​

Option A — Receipt only on connectivity probes ​

Detect Hinata-probe shape in user message ("connect", "are you connected", "hinata?", /whoami). When detected, append the receipt footer. Otherwise reply normally without footer noise.

Add to hinata-bot-poller.py::reply_in_voice:

python
PROBE_PATTERNS = re.compile(r'\b(connect|connected|hinata\?|whoami|status)\b', re.IGNORECASE)
if PROBE_PATTERNS.search(msg):
    # Append vault-receipt footer to system prompt instruction
    system_prompt += "\n\nThis message is a connectivity probe. Include a vault receipt in your reply: state your name, which context.md you loaded, and a one-line summary of Michael's current focus in your domain."

Option B — /whoami slash command ​

Bot exposes /whoami as an explicit Telegram command that returns the receipt. Doesn't interfere with normal replies. Cleaner separation.

Option C — System-prompt addendum (always-on) ​

Every reply includes a receipt. Verbose; probably too much for short answers.

Recommendation: B for explicit probe, A as a smart fallback for ambiguous probes.

Pickup work ​

  1. Add /whoami CommandHandler to hinata-bot-poller.py per Option B.
  2. Add PROBE_PATTERNS detection per Option A.
  3. Add last_touched field to each Commander's context.md frontmatter (some have it, some don't — Jimmy Neutron sweeps weekly).
  4. Test on each of the 6 bots — message /whoami to each, verify receipt is correct + cites the right Commander.

Source signals (now in inbox/done/):

  • 3 inbox files where Michael probed connectivity and got insufficient proof.