team-telnyx/ai / providers/claude/plugins/telnyx-ai/skills/telnyx-meeting-bot
telnyx-meeting-bot
VerifiedUse when an agent must join, observe, react in, transcribe, summarize, or follow up on a Zoom, Google Meet, Microsoft Teams, or Webex meeting with Telnyx Meeting Bot. Handles vague requests, request-specific live polling, name/phrase and semantic triggers, explicitly authorized speak/chat actions, recovery, and all implemented transcript artifact types.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 29.1 KB |
| references/artifact-selection-and-recovery.md | markdown | 6.6 KB |
| references/repeating-semantic-actions.md | markdown | 4.3 KB |
references/repeating-semantic-actions.md
View on GitHub# Repeating Semantic Action Protocol
Use this protocol only when the requester explicitly asks a semantic `speak` or
`send_chat` rule to repeat. One-shot rules use the simpler session-and-rule claim
key in the parent skill.
## Persisted Rule State
Persist one record scoped by `(session_id, rule_id)`:
```json
{
"lease_owner": null,
"lease_expires_at": null,
"generation": 0,
"last_evaluated_seq": 0,
"context_segments": 3,
"active_occurrence": null
}
```
When active, `active_occurrence` is:
```json
{
"occurrence_first_seq": 41,
"evidence_seqs": [41, 42],
"claim_key": "action:<session_id>:<rule_id>:repeat:41",
"status": "active"
}
```
Persist the semantic condition, action payload, classifier/configuration, and
fixed `context_segments` with the rule. Do not change them while the rule is
active.
## Ordered Evaluation and Canonical Evidence
1. Acquire a durable exclusive lease for `(session_id, rule_id)` with an atomic
compare-and-swap (CAS). Only the current lease owner may classify transcript
windows or update this rule state. A replacement worker must wait for lease
expiry and then acquire a newer `generation`.
2. Consume finalized transcript segments in strictly increasing `seq` order.
Evaluate each sequence once using the fixed trailing window ending at that
sequence, but do not advance the durable cursor in a separate write.
3. When no occurrence is active and the condition becomes true, choose the
**shortest contiguous suffix ending at the current sequence** that still
satisfies the persisted semantic condition. This structurally determines one
evidence list; its first element is `occurrence_first_seq`. If semantic
classification is stochastic, only the lease owner runs it and persists the
selected evidence through the transaction below—other workers never
independently classify a committed sequence.
4. For every evaluated sequence, execute one per-sequence CAS transaction that
compares the prior `generation`, lease owner/validity, and prior
`last_evaluated_seq`. That same transaction must advance `last_evaluated_seq`,
persist the evaluation result and canonical `evidence_seqs`, and perform the
applicable occurrence transition: create `active_occurrence` plus the durable
`claimed` action `action:<session_id>:<rule_id>:repeat:<occurrence_first_seq>`,
retain the active occurrence, atomically clear it, or retain no occurrence.
5. After that transaction commits, dispatch only if a second CAS changes that
claim from `claimed` to `dispatching` immediately before transport invocation.
A CAS loser reloads state without dispatching.
6. While an occurrence is active, every positive overlapping window reuses its
persisted claim key. It cannot create another claim, even if the newest
evaluation sequence differs or a shorter evidence suffix later appears.
## Clearing and Re-arming
Clear `active_occurrence` only when a later ordered evaluation satisfies both:
- its fixed context window contains none of the persisted `evidence_seqs`; and
- the persisted semantic condition evaluates false.
The clear is the occurrence transition inside the same per-sequence transaction
that advances `last_evaluated_seq`; never advance the cursor and clear in separate
writes. Compare the current `generation`, lease owner/validity, active claim key,
and prior cursor. An older/stale worker must fail that CAS and reload without
clearing. A new repeat claim is permitted only after the clear commits and a
still-later ordered evaluation produces a new false-to-true transition.
## Crash Recovery
- Resume from the persisted `last_evaluated_seq`; its evaluation and occurrence
transition committed in the same transaction, so never reclassify it.
- If the lease expired, acquire a newer generation before evaluating.
- If `active_occurrence` exists, reuse its claim key and action state.
- A recovered `claimed` action may compete once for the `claimed` → `dispatching`
CAS. A recovered `dispatching` action becomes `outcome_unknown` before any
trigger evaluation unless durable transport evidence proves no bytes were sent.
- Preserve the parent skill's side-effect rule: never automatically repeat an
accepted or outcome-unknown `speak`/`send_chat` dispatch.
This protocol intentionally favors at-most-once action delivery over reacting to
ambiguous consecutive utterances that never produce a clear transition.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |