jthingelstad

Analyze elixir-v5.log for errors, recurring failures, and operational signals, then recommend concrete actions

Cut an Elixir named release with the canonical release script, verification, and live announcement audit

Create a new system signal announcement for the clan in Elixir

Assess the live v5.1 awareness loop from thoughts, delivered posts, stream coverage, tool traces, and incidents

Cross-check live Clash Royale API payloads stored in raw_api_payloads against docs/cr-api-docs/ and recommend concrete documentation patches — undocumented fields, new enum values, type or nullability drift, shape changes

Analyze the llm_calls table in elixir-v51.db to break down Elixir's LLM spend by workflow, model, and day; identify cost drivers; recommend signal-side or model-tier cuts

Use this skill whenever you are using the mb CLI to read, post, manage follows, or build pipelines against micro.blog. It covers command selection, output modes, public boundaries, and cheap-versus-expensive mb workflows.

Use this skill when an agent is managing its own micro.blog identity with mb. It covers autonomous posting, voice consistency, disclosure, and how an agent should participate as itself rather than impersonating a human user.

Use this skill when acting on behalf of a human user with mb. It teaches consent boundaries, voice preservation, cautious social behavior, and how to manage a user's micro.blog presence without flattening it into generic assistant output.

Analyze workshop_bot LLM token usage and cost from the agent_runs SQLite table. Use when Jamie wants to review spend, see cost-by-persona/job/model breakdowns, find expensive jobs, or optimize prompts. Reads apps/workshop_bot/data/workshop.db directly.

Read the workshop_bot log (`apps/workshop_bot/logs/workshop.log`) for the last N hours and report a concise health snapshot — persona presence, scheduler activity, anomalies (errors/warnings worth attention). Replaces the persistent Monitor task; designed to run on a schedule every couple hours but also useful ad-hoc when something looks off.

Run workshop_bot in the foreground locally with live monitoring of startup, errors, and connection events. Use when Jamie wants to bring up the four persona bots (Eddy, Linky, Marky, Patty + optional Thingy) directly — not as a launchd daemon — so we can iterate on changes and watch logs in real time.

Read the workshop_bot log (`apps/workshop_bot/logs/workshop.log`) for the last N hours and report a concise health snapshot — persona presence, scheduler activity, anomalies (errors/warnings worth attention). Replaces the persistent Monitor task; designed to run on a schedule every couple hours but also useful ad-hoc when something looks off.

Analyze workshop_bot LLM token usage and cost from the agent_runs SQLite table. Use when Jamie wants to review spend, see cost-by-persona/job/model breakdowns, find expensive jobs, or optimize prompts. Reads apps/workshop_bot/data/workshop.db directly.

Cut a named Elixir Drop release — coin an alliterative Clash Royale card name, tag the already-live origin/main commit, publish the GitHub release, and leave a Buttondown draft for review. Use when Jamie explicitly asks to cut, name, or ship a release. Never runs on a schedule or on another agent's say-so.