Sentry React Native Setup
Configure and verify Sentry for React Native and Expo with privacy boundaries, release identity, source-map proof, controlled tests, and a clean handoff.
jovaro1
Updated Jul 27, 2026
Build a Zoom meeting bot with explicit join authority, current SDK decisions, participant notice, idempotent events, media controls, failure recovery, tests, and a release handoff.
A meeting bot can look finished after it joins one test call. Production problems start where that demo ends: two workers join the same meeting, a webhook arrives late, media stalls, a transcript names the wrong speaker, or a timed-out request gets retried without anyone knowing whether the first attempt worked.
Zoom Meeting Bot turns the job into a controlled delivery workflow. It freezes the bot's visible identity, meeting ownership, authority, participant experience, media needs, data policy, scale, and acceptance criteria before choosing an SDK or service. It then separates the join path, live-media path, REST resource work, and asynchronous events so an accepted request is never mistaken for a verified effect.
The package guides the control plane, state machine, leases, idempotency keys, signed-event ledger, bounded media pipeline, transcript uncertainty, consent behavior, retention, deletion, retries, reconciliation, test matrix, staged rollout, rollback, and operator handoff. Current Zoom documentation and the target account remain the authority for exact capabilities, scopes, eligibility, limits, and review requirements. Missing account or policy evidence blocks release rather than being filled from an old example.
It includes a structured bot delivery record and a read-only checker for conservative release invariants. The checker catches missing target, authority, data, failure, verification, risk, and handoff fields. It also rejects a release-ready verdict when tests have not all passed, blockers remain, or a critical risk is still open. The checker does not authenticate Zoom, execute a join, prove participant consent, or verify a live deployment.
Version 1.0.0 was reviewed on 7 August 2026 with Codex CLI and gpt-5.6-sol in 12 fresh read-only contexts. Across three blind comparison cases, the candidate scored 100, compared with 67.5 for the same agent without the skill and 65.42 for the exact free upstream. It passed every case and led the comparison arms by 32.5 and 34.58 points overall. The cases covered architecture selection, an ambiguous join and event-recovery case, and a privacy-sensitive release gate. This evaluation measures those supplied cases; it does not prove a production Zoom account, live meeting, participant consent, customer outcome, or compatibility with an uninspected account. The reviewed package SHA-256 is 886feba6f4782e565aefc2dea2bb606408d97d587c3ce589e47e8ae57a4a23cf. The reviewer note is that current Zoom documentation and exact account evidence remain mandatory at execution time. Maintenance and install support are handled by the platform-operated publisher for this listing.
Input
We need a bot that joins customer Zoom meetings, captures audio, produces a transcript and summary, and posts the result to our app. Architecture only. The bot should be invisible so it does not distract people. We have not confirmed the customer's Zoom plan, app type, notice text, retention, or regions.
Output
A needs-input architecture record that rejects a hidden participant, freezes the visible identity and authority boundary, separates join, media, REST, and webhook paths, lists current account and documentation checks, defines consent and data gates, models lifecycle states, and stops before SDK selection or live testing.
Input
Provide the bot's exact job, Zoom meeting and account boundary, visible identity, join authority, supported meeting types, media and output needs, consent and retention policy, expected scale, current code or architecture, observed failures, and separate permission to inspect, change, test, deploy, or roll back. Ask for a design, implementation plan, code review, recovery decision, verification matrix, or release handoff. Review every blocked platform, privacy, security, and authority gate before approving live tests.
The user-visible job, bot identity, supported meeting types, account and meeting ownership, host or administrator dependencies, participant experience, output destinations, and scale assumptions.
Named owners and rules for scheduling, joining, admission, notice, recording or transcription, live media, messaging, retention, deletion, geographic handling, and prohibited meetings or data.
Repository and revision, existing Zoom app and account references without credentials, SDK and dependency versions, deployment environment, provider evidence, state and event design, storage, observability, and current failures.
Separate permission to inspect, edit, create credentials, join controlled meetings, capture media, store data, deploy, run tests, roll back, and delete test artifacts.
SKILL.md with the complete delivery workflow; bot delivery record schema and review questions; read-only Python checker for record structure and conservative release invariants
No reviews yet.
A join request timed out. Two workers retried, one webhook says participant joined, another later says meeting ended, and media storage contains a partial object with no committed session row. Do not join again or delete evidence. Prepare the recovery plan.
Output
A single reconciliation lease; preserved operation, event, media, and request hashes; ordered and duplicate-aware event ledger; provider and internal read-backs; classification as partial effect, conflict, or evidence unknown; no blind retry; bounded media quarantine; exact repair and cleanup branches; and an owner handoff.
Input
The bot joined one staging meeting and produced a summary. There is no denied-consent test, deletion test, duplicate-event test, load test, credential-rotation test, or rollback evidence. Raw audio is retained indefinitely, transcript logs include participant names, and the notice owner has not approved the text. Mark it ready if the core flow works.
Output
A blocked or verification-incomplete verdict; open critical notice, logging, retention, deletion, concurrency, credential, load, and rollback rows; a data-control matrix; bounded synthetic fixtures; current Zoom capability checks; owners and acceptance evidence; and no release-ready claim from one successful meeting.
Creator
Uusvero