tkellogg

Resolve-and-promote security gate — verifies a Telegram (or other in/dm/*) chat id against the phonebook and, only when it resolves to owner, promotes it onto in/human/owner. Fail-closed; a stranger's message stops at an unresolved sighting.

Telegram bridge (daemon) — getUpdates messages become in/dm/telegram/<chat.id> conversation ingress; in/package/telegram/send delivers replies with sender=telegram receipts. A transport is just a package (zero kernel edits).

Runbook for diagnosing common lanius RUNTIME / operational failures — broker & bus errors ("CONNECT refused", "bad credential"), stray/orphaned processes, ports stuck in use, the daemon not dispatching, the web UI not loading. Use when something is failing at runtime (not a compile error or a logic bug) and you want the symptom → cause → diagnose → fix → prevent steps. Each file in this skill is ONE runbook entry, keyed by the symptom you actually see. Start here, match your symptom to an entry, follow it.

The journey → handoff → implement → verify pipeline for shipping substantial work on lanius, with deliberate model-tiering across the phases. Use when Tim asks to plan a feature, write or "knock out" handoff docs, implement a spec, or orchestrate implementation + verification — anything bigger than a quick edit. Encodes who plans vs implements vs verifies, which model/effort fits each phase, how to dispatch clean-context worker agents, and the containment + commit discipline that keeps the repo clean. Fable should *ALWAYS* follow the second-level workflow specified below.

How agents talk on Lanius - to the human (send_message vs ask_human), to coding workers (deliver/spawn/inbox), and to native/profile agents (agent catalog/run/spawn), when to speak unprompted vs stay quiet, when to set priority, shared-channel etiquette, and the failure-mail contract. Read before messaging the human, dispatching work to another agent, launching a native/profile agent, or coordinating in a shared room.

The chat/conversation view — owns the chat protocol on the bus and answers the chat-shaped conversation-list + introspection queries over local HTTP from the sqlite truth. Requires the history package.

The model-tiering knowledge base — which model to use for planning, implementation, and verification. Read a kb/role-*.md file when you are about to dispatch or pick a model for a task.

Read, search, and write knowledge bases (kb/ folders). Use when you learn something durable that is "ours" (not just your own scratch notes) — model tiering, an API's quirks, a project convention — or when you need a fact someone already wrote down.

How to choose between markdown, HTML messages, and UI plugins in the local web chat — and how to render each well. Use whenever you are about to call `send_message` to a web-chat channel, especially when the content has structure (rows, columns, statuses, a state machine, multiple linked things) or when the right answer is "let me show you", not "let me tell you". Do not use for Discord channels — HTML is rejected there.

Create and manage runtime command hooks declared in skills/*/hooks.json. Use when the user asks to intercept tool calls, augment prompts, audit or mutate tool arguments/results, add startup/shutdown scripts, or debug hook behavior.

Index of recurring problem shapes — environmental awareness (world-scanning, OS event hooks, browser automation), communication (messaging, multi-agent handoff, async waits), coordination between parallel things (avoiding collisions, oscillation, double-work), and self-correction (circuit-breaker, try-harder). Use when designing a non-trivial workflow, when the same shape just happened twice, or when the answer to "what should I do here?" isn't obvious. Don't read on every turn.

Structured root cause analysis for arriving at a concrete action. Use when something went wrong, a pattern keeps recurring, behavior has drifted, or you catch yourself resolving to "do better" / "remember to X" without a concrete artifact. Five-whys forces behavioral resolutions into file edits, config changes, memory block updates, or scheduled jobs — the action item must produce a diff someone else can verify. Do not use for simple debugging with an obvious cause.

Diagnose agent behavior using event logs, journal history, and scheduler state. Use when you need to understand why something went wrong, review your own patterns, audit scheduled jobs, or debug communication issues. Do not use for one-off messaging or memory management (use the memory skill instead).

Guide for the first days with a new human. Use when: the `init` memory block exists (pointing you here), you have no persona/communication blocks yet, or a human explicitly asks about setup. This skill is about having good conversations — not filling out forms.

Run shell commands in the background with file-backed output capture and a completion event that wakes the agent when the job exits. Use when a command might take more than ~30 seconds (builds, tests, deployments, agent jobs like acpx/codex exec) and you want to keep working while it runs.

Evaluate whether journal predictions became true 2-3 days later using event and Discord evidence, then log structured outcomes for calibration and behavior updates. Use when auditing prediction quality. Do not use for one-off messaging tasks.

Read-focused navigation of Bluesky for book research — search users and posts, view a user's feed, fetch a specific post and its replies, search within an author's posts. Use whenever you need to look up what someone said on Bluesky, find quotes, gather context on a handle, or understand a thread. Posting section included but requires PDS credentials (not configured in this repo).

This skill should be used when the user asks to "generate an image", "create a diagram", "make a chart", "visualize data", "create a cover image", or when a blog post needs visual content like mermaid diagrams, matplotlib/seaborn charts, or AI-generated images via Google Gemini.

Use when the user wants to add a React/JSX component to the Jekyll site as a page. Triggers on "add jsx", "put this jsx on the site", "render this react component", or when given a .jsx file to integrate.

Review a blog post draft for readability, interestingness, and factuality. Use when Tim asks to "review", "critique", "check", or "give feedback" on a post. Applies Tim's writing style standards and pushes for provocative, scannable, memorable writing.

This skill should be used when the user asks to "write a blog post", "draft a post", "write in my style", "help me write", or when creating or editing content for the Jekyll blog at tkellogg.github.com. It encodes Tim Kellogg's distinctive writing voice and should guide all blog content generation.

Select and compare language models using Artificial Analysis benchmark scores, pricing, speed, and task fit. Use when choosing a model for coding, writing, knowledge work, agents, math, science, reasoning, multilingual, or multimodal tasks; when comparing cost per intelligence; or when determining which model names and aliases are available in the current Codex or Claude Code runtime.