1 seat, 1 active in the last 7 days. These are people on the organisation's Claude account. Whichever surface a seat worked in, its usage lands here: seats, not separate agents.
Skills library
JamesWeb3/sentry-aios ↗
The versioned workflow skills these seats run on, shipped through the organisation's plugin marketplace.
| Seat | Model | Calls | Tokens | Tools | Share | Last seen |
|---|---|---|---|---|---|---|
| james.oldham0604@gmail.com | claude-opus-562%claude-opus-5[1m]29%claude-sonnet-54%+3 more | 1011 | 449.1M | 827 | 100% | 7 Aug |
Loads counted from telemetry, all time · badge shows total loads
sentry-atlas
0Proposal & pitch deck production using Google Slides. Use when the user asks for a proposal, SOW, pitch deck, sales brief, or "put together a deck" for a specific client or opportunity. Always works template-first — clones a master Slides template, fills its `{{token}}` placeholders, optionally inserts client logo, and shares the result.
not used yet
sentry-branding
0Canonical Sentry AI brand system — colours, type, components, voice, motion, and iconography. Read before touching any Sentry-branded surface (sentry-ai marketing site, sentry-aios platform, sentry-ui block library, any client-facing Sentry collateral). The single source of truth that keeps every Sentry repo visually and tonally consistent.
not used yet
sentry-create-ticket
0Turn a conversational request in chat into ONE clear, well-structured engineering ticket. Interview just enough to fill the ticket frame (title, client, repo, priority, context + acceptance criteria), show it back to confirm, then create it with tickets_create assigned to Sentinel. Use when James types /create-ticket or asks to "make a ticket" from a chat.
not used yet
sentry-curator
0Knowledge-Graph maintenance. Two modes — "advise" (review the latest wiki-lint run + delta, write 2-3 prioritised activatable suggestions) and "execute" (perform one approved suggestion: mechanical fixes directly, judgement calls as a PR/approval). The Karpathy self-learning wiki loop for the Sentry vault.
not used yet
sentry-engineer
0Autonomous engineering workflow — clone a client repo, make code changes, run typecheck/lint, commit, push a feature branch, open a pull request. Use when the user asks for a feature, bug fix, refactor, or any code change that should ship as a PR. Always works on a fresh clone in a tmp directory and cleans up after itself, so there's no chance of stale state or out-of-sync workspaces.
not used yet
sentry-pseo-article
0Write and publish ONE buy-intent pSEO article from the aios_pseo_keywords queue — claim the top-scored keyword, write a GEO+SEO formatted article with a quotable summary, FAQ from real Search Console questions and one proprietary statistic, publish to the parasite platform (GitHub ai-operations-guides), verify the URL is genuinely public, update the row. Use when the nightly pSEO routine fires or James asks to publish a pSEO article.
not used yet
sentry-pulse
0Marketing performance analysis — answer questions about Meta Ads (spend, CTR, CPM, CPC, conversions) and Google Analytics 4 (sessions, top pages, traffic sources, realtime users). Use when the user (or a Slack channel binding) asks how ads or the site are performing, requests a pulse, drills into a specific campaign or page, or wants a week-over-week comparison. Read-only — never mutates campaigns or property settings.
not used yet
sentry-scribe
0Meeting transcript processing — take a single inbound Fathom transcript, write a vault meeting note, log to client dossiers, create engineering tickets for follow-ups, and draft proposals when the call asks for one. Use when fired by the Fathom webhook with an `aios_meeting_events` row to process. Always processes exactly one meeting per fire and marks it processed when done.
not used yet
sentry-sentinel-loop
0The Sentinel loop — turn a Slack message in #sentinel into a ticket, ship the fix as a PR via the sentry-engineer runbook, verify it with a subagent, post the Vercel preview link + a queued merge approval back into the thread, and (after James replies "ship it") the platform merges, deploys, and you confirm live. Use whenever a request arrives from the #sentinel Slack channel.
not used yet
sentry-ship
0The submission gate every code change passes before it becomes a PR. Takes a finished branch and runs the full pre-PR discipline — clean the branch, re-prove typecheck/lint, self-audit the diff with fresh eyes, run a security review, then commit and open a neatly-formatted pull request (and NEVER merge). Use at the end of any coding task, right before opening a PR, in any repo. sentry-engineer and sentry-sentinel-loop hand off to this skill to open the PR.
not used yet
sentry-subagent-pattern
0How to dispatch focused subtasks to a specialized subagent without bloating your own context. Use when you need work done by a different specialist (Quill firing Atlas for a proposal), when a task is long-running, or when you want parallel work across multiple subtasks. The subagent runs to completion in isolation, you read a short summary when done.
not used yet
sentry-triage
0Claim and execute the oldest pending routine run. Called by sentry-drain on every /triage hit. Discovers pending work via runs_list_pending, claims one run atomically, follows the routine's embedded instructions, then marks it complete or failed.
not used yet
sentry-vance
0Weekly SEO blog engine for the Sentry AI website. Use when running the "Weekly SEO Blog Engine" routine (fires every Sunday) or when James asks for a new SEO blog article. Researches a trending, high-intent topic, drafts one publish-ready article into the website repo's blog data, opens a pull request on JamesWeb3/sentry-ai, and notifies James to merge. Drafts and proposes only — never merges, never publishes directly.
not used yet
sentry-watch
0The WATCH orchestrator — Sentinel triages a ticket into the right starting phase and runs it through the pipeline (Workshop → Architect → Task → Check → Hand off) on its own, advancing the ticket's watch_phase as it goes and pausing for James's review only when it genuinely needs a human. Use this whenever Sentinel picks up a ticket; it decides which phase skills to run and in what order.
not used yet
sentry-watch-architect
0WATCH phase 2 (Architect) — Sentinel turns the Workshop brief into a concrete, proven build plan. Writes task.md (the step-by-step strategy the build follows) and test.md (how it will be verified) at the repo root. No feature code yet. Assigned to Sentinel when a ticket's WATCH phase is set to Architect.
not used yet
sentry-watch-check
0WATCH phase 4 (Check) — Sentinel's audit gate. Runs test.md, typecheck/lint, the Sentry compliance + security checks, and `npx react-doctor@latest` across the code Task produced, cross-checks against the acceptance criteria, and logs findings. Fixes or kicks back to Task if anything fails. Assigned to Sentinel when a ticket's WATCH phase is set to Check.
not used yet
sentry-watch-handoff
0WATCH phase 5 (Hand off) — Sentinel opens the PR with an in-depth description (what/why, the task.md strategy, test.md + Check results, preview URL, acceptance-criteria checklist, risks), then watches the GitHub automations/CI on the PR and reports status. Never merges. Assigned to Sentinel when a ticket's WATCH phase is set to Hand off.
not used yet
sentry-watch-task
0WATCH phase 3 (Task) — Sentinel executes the build. Follows task.md closely, writes the code and the tests from test.md via the sentry-engineer discipline (fresh clone, branch, tsc + lint), pushes the branch, then moves the ticket to Review and notifies James. No PR yet — that's Handoff. Assigned to Sentinel when a ticket's WATCH phase is set to Task.
not used yet
sentry-watch-workshop
0WATCH phase 1 (Workshop) — Sentinel's planning/intake step. Take the ticket plus whatever James points you at (usually a client folder with profile.md + actions.md), read the profile, organise the problem into a clear brief + acceptance criteria, and log what you're doing into actions.md. Pure information organisation — no task.md, no code. Assigned to Sentinel when a ticket's WATCH phase is set to Workshop.
not used yet
| Tool | Runs | Seats | Rejected | Failed | Avg |
|---|---|---|---|---|---|
| Bash | 425 | 1 | – | – | 7.8s |
| Edit | 177 | 1 | – | – | 0.0s |
| mcp_tool | 89 | 1 | – | – | 1.2s |
| Read | 57 | 1 | – | – | 0.1s |
| Write | 41 | 1 | 1 | – | 0.0s |
| ToolSearch | 13 | 1 | – | – | 0.0s |
| WebFetch | 11 | 1 | – | – | 7.8s |
| TodoWrite | 9 | 1 | – | – | 0.0s |
| Skill | 3 | 1 | – | – | 0.0s |
| WebSearch | 2 | 1 | – | – | 12.3s |
425 runs across 1 seat · 7.8s avg
| Command | Runs | Seats | Failed | Avg |
|---|---|---|---|---|
| unknown | 425 | 1 | – | 7.8s |
Most recent