teleo-infrastructure/.agents/skills/leo-telegram-canary-ops/SKILL.md
twentyOne2x 15c30f1fbe
Some checks are pending
CI / lint-and-test (push) Waiting to run
Make Leo skill onboarding clean-context reproducible (#150)
* make Leo skill onboarding clean-context reproducible

* harden Leo clean-context skill acceptance
2026-07-15 03:19:15 +02:00

3 KiB

name description
leo-telegram-canary-ops Use for live Leo Telegram tests through authenticated Chrome/Computer Use, including memory, KB audit, staged DB writes, screenshots, and clear separation from external outreach.

Leo Telegram Canary Ops

Job

Run Telegram-visible Leo canaries when testing Leo itself, with retained screenshots and DB readbacks.

Trigger Phrases

  • "test Leo in Telegram"
  • "send the Leo group message"
  • "Telegram-visible proof"
  • "Chrome Telegram canary"
  • "Leo remembers conversation"
  • "staged write through Telegram"

Authorization Boundary

Live Telegram bot messages to the Leo group are authorized when they are test canaries for Leo behavior.

This is different from external outreach. Do not send external outreach, partner messages, public announcements, or non-test communications without exact authorization.

Required Tooling

Use Computer Use through node_repl and the Chrome/Computer Use skill path. Do not use:

  • AppleScript,
  • osascript,
  • JXA,
  • System Events,
  • focus hijacking,
  • foregrounding hacks.

Canary Types

Name the canary before sending:

  • bot readiness,
  • memory,
  • KB audit truth,
  • proposed-vs-applied truth,
  • open-ended m3taversal-style triage,
  • staged write,
  • no-canonical-mutation proof,
  • screenshot/readback proof.

Message Discipline

Use unique markers such as WL-LIVE-TG-M3TAVERSAL-YYYYMMDD-Tn.

Ask Leo for machine-checkable reply markers like:

  • LIVE_TG_TURN1_ACK
  • LIVE_TG_TURN3_STAGED
  • LIVE_TG_TURN4_READBACK

Do not rely only on exact-ID prompts. At least one representative canary should be a vague operator-style prompt such as: m3taversal says the agents are not working, the KB is in the same state as last night, and Leo should be able to manipulate the KB; ask Leo what that means, what it would inspect first, what fixed means, and how it separates proposed, approved, and applied. Keep this canary read-only unless the task explicitly authorizes a staging or apply step.

Required Proof

After each live Telegram canary:

  1. Capture screenshot.
  2. Record exact message sent.
  3. Record Leo reply marker and substance.
  4. Query DB if the canary involves DB state.
  5. Record canonical public table counts when proving no mutation.
  6. Save a local markdown/json artifact.
  7. Sync artifacts to VPS report dir when relevant.

Current Known Proof

  • Live memory and KB audit are retained in docs/reports/leo-working-state-20260709/telegram-live-canary-current.json.
  • Live staged-write proof is retained in docs/reports/leo-working-state-20260709/telegram-live-db-write-canary-20260709.json.
  • Open-ended m3taversal-standard read-only triage is retained in docs/reports/leo-working-state-20260709/telegram-live-open-ended-suite-current.md.

Historical receipts contain immutable legacy marker IDs. Treat them only as evidence identifiers; never reuse them as a participant name or future marker.

Use docs/reports/leo-working-state-20260709/current-truth-index.md for artifact paths.