bugfix-dispatcher
Protocol and launcher for running the bugfix-dispatcher resident session. Routes incoming bug reports and feature requests from other Claude Code sessions into isolated builds on easypost-sandbox repos, independently verifies against merge bars (tests, code review), and auto-merges or opens PRs. Use when maintaining a long-lived dispatcher-role session via `bin/bugfix-dispatcher-launch`. Don't use for one-off builds or repos outside easypost-sandbox.
claude-tail
Analyzes Claude Code session logs (JSONL format) with syntax highlighting, filtering by event type/tool/time/errors, and real-time following. Use when debugging a session, monitoring tool usage, understanding workflows, or tracking errors. Don't use for reading general Claude Code documentation or for non-session-log files.
data-pipeline
Data pipeline architecture patterns and best practices, including medallion/three-layer architecture (Raw/Staging/Enriched or Bronze/Silver/Gold), YAML-based schema management, and ETL workflow patterns. Use when designing or implementing data pipelines, working with data warehouse layers, or managing table schemas in YAML.
dispatcher
Generic protocol for a resident work-dispatcher session attached to a repo — receives change requests via cross-session messages, builds each in an isolated background worker session, independently verifies the result, and merges, opens a PR, or applies per a per-repo config. Generalizes recover-on-start, per-resource locking, a global WIP ceiling, launch verification, edge-triggered idle subscription, a mechanical merge/apply bar, and safe cleanup, as proven in bugfix-dispatcher, infra-dispatcher, and matrix-dispatcher. Use when attaching an autonomous work-dispatcher to a repo receiving change requests via cross-session messages, or standing up a new domain-specific dispatcher. Don't use for a one-off task with no recurring request stream, a repo that doesn't want autonomous merges/PRs/applies, or to modify bugfix-dispatcher, infra-dispatcher, or matrix-dispatcher themselves — those are independent, already-in-production skills.
distill
Reduce a document to its minimum effective dose — the least information needed to fully convey the purpose and key concepts. Use when asked to simplify, condense, distill, or strip a document down to essentials. Don''t use for paraphrasing or adding explanation. Triggers on: "simplify this", "distill this", "trim this down", "what''s the minimum here", "make this shorter without losing anything important".
duckdb
Use DuckDB for analytical SQL over local CSV, JSON, Parquet, and Excel files - aggregations, joins, window functions, statistics, and data profiling without an import step.
gcp
Google Cloud Platform patterns and quirks — general GCP infra (IAM roles, Cloud Run deploy mechanics, Secret Manager, Cloud Monitoring) plus Vertex AI specifics for Claude/Anthropic models. Use when working with GCP infrastructure generally (IAM, Cloud Run, Secret Manager, monitoring), not only when Vertex AI or Claude-on-Vertex is involved. For OpenTofu/Terraform resource authoring see the opentofu skill; for GitHub Actions CI/CD auth and deploy workflows see the github-actions skill; for BigQuery query cost discipline see the bigquery skill — this skill covers the GCP service-level behavior those skills build on top of.
git
Git version control conventions and GitHub CLI workflows including commits, branches, pull requests, and code reviews with professional message practices. Use when managing commits, branches, PRs, or coordinating multi-worktree workflows. Don't use for deployment scripting or infrastructure-as-code (use other domain skills instead).
github-actions
GitHub Actions workflow design for deploying to GCP — Workload Identity Federation (keyless OIDC), per-resource least-privilege IAM, path-filtered multi-workflow layouts, and digest-based Cloud Run deploys. Use when creating or editing .github/workflows/*.yml that build/push images or deploy to GCP, when setting up CI auth to a cloud provider, or when a workflow needs to trigger on specific file paths for one of several deployable components in the same repo.
go
Go development conventions including gotestsum for testing, standard tooling (gofmt, go vet, -race flag), and best practices for the Go environment. Use when developing, testing, or debugging Go code. Don't use for deployment, containerization, or Rust/Python projects.
javascript
Node.js and JavaScript development patterns, gotchas, and library compatibility notes. Use when working with Node.js projects, TypeScript, React, antd, Zustand, Playwright, Bun, or Vitest.
josh-email-voice
Writing style and workflow conventions for drafting email as Josh Lane (CTO, EasyPost). Apply whenever drafting a Gmail reply or note on his behalf. Don't use it for Slack messages (use josh-slack-voice instead), formal HR/legal documents, or communications from EasyPost as an organization.
jq
JSON processing with jq for filtering, transforming, aggregating, and reshaping data from files, APIs, and command output on the command line. Use when processing JSON via command-line filters. Don't use for CSV/TSV (use xsv), YAML (use yq), or analytical queries requiring SQL (use duckdb).
just
Task automation with just and justfiles for recipe definition, dependencies, parameters, imports, and shell script delegation. Use when building new task runners or maintaining justfiles. Don't use for existing make-based projects or arbitrary shell script work.
looker-development
Develop LookML against a live, GitHub-CI-integrated Looker instance (the official "Looker Continuous Integration" GitHub App plus git-connected deploy). Use when debugging `looker/ci - Github CI - *` GitHub check failures or silence, setting up Looker CI (GitHub App install, CI Suite creation, deploy webhook) for a repo that doesn't have it yet, onboarding a new LookML model, diagnosing a connection/SQL Validator failure, or trying to shift feedback left before the multi-minute CI round-trip. Not for querying/searching an existing Looker catalog for data (see the `looker` skill) or for BigQuery cost/query mechanics (see the `bigquery` skill).
matplotlib
Figure engineering patterns, gotchas, and reusable helpers for matplotlib. Use when directly building standalone figure scripts, diagram generators, or canvas visuals outside the epq render pipeline — e.g., generate_layer_icons.py, vision project banners, morning-brief charts, competitive analysis figures. Covers shape boundary math, arrow/edge geometry, series consistency, font registration, zorder, Unicode rendering, and canvas clipping. Distinct from the epq skill (which covers the Quarto/LaTeX analysis pipeline). Load alongside epq when figure work involves both the pipeline and standalone diagram construction.
methodology
Consolidated reference for phased execution (continuous work through genuine blockers only, adaptive todo granularity), visual communication standards (Mermaid, formatted tables, LaTeX instead of raw data), and verification tiers for code changes. Load when working through multi-step tasks needing phased-execution guidance or designing test/review gates for a code change. Don't use for project-specific tool tutorials or EasyPost domain procedures — refer to domain-specialized skills instead.
operating-lessons
Situational verification, delegation, and coordination rules: sub-agent output trust (numeric claims, data availability, infra, hallucination), peer-session coordination, receipt-contract agents, worktree/briefing gotchas, interactive-tool/auto-approve pressure, search/1Password gotchas, and technical corrections (BigQuery join fanout, replace_all safety, interactive Bash prompts, worktree cleanup order). Use before messaging a peer session or asking Josh a scoping/sequencing question; reviewing sub-agent output before surfacing it; briefing an Agent call or choosing a receipt-contract agent; targeting, finishing, or merging a worktree-isolated branch (including an orchestrator git refusal on the shared checkout); running an apply or interactive tool under auto-approve; searching for an internal company document; or double-checking a verification/safety procedure from CLAUDE.md. Don't use for general TDD, git commit conventions, or domain gotchas covered by git, javascript, python, or bigquery skills.
org-announce
Draft org-level or company-wide announcements (reporting changes, promotions, departures, strategy, business decisions). Produces email and Slack formats in one invocation. Enforces Josh Lane's register: terse, declarative, no sentiment, no signature, no sycophantic close. Use when asked to draft an announcement, org update, or leadership communication.
presenterm
Build and run terminal-based presentations from markdown with presenterm - slide syntax, speaker notes, executable code blocks, mermaid/d2/LaTeX rendering, themes, and PDF/HTML export. Use when building a presentation from markdown slides with live code execution or diagram rendering. Don't use for Beamer/Reveal.js decks, browser-based slide tools, or EasyPost branded presentation decks — use ep-slides for those instead.
python
Use uv for fast Python project management, script execution, dependency handling, and tool installation. AVOID pip - always use uv commands (uv add, uv sync, uv run) instead.
qmd-math
Math notation conventions for Quarto/EPQ documents rendered via lualatex. Use when: writing or adding a formula, equation, or mathematical expression to a .qmd file; asked about display math, inline math, or LaTeX notation in a QMD/Quarto context; defining a where-clause or variable definitions for an equation; converting prose variable descriptions into structured math notation; fixing math that renders badly in a PDF; using \lvert, \begin{aligned}, \tfrac, \text{where}, or any LaTeX math command in a Quarto document. Triggers on: "write a formula", "add an equation", "math notation", "where clause", "display math", "LaTeX equation", "define variables", "simplify the formula", "typeset this".
quarto
Render computational documents to markdown (DEFAULT), PDF, HTML, Word, and presentations using Quarto. PREFER markdown output for composability. Use for static reports, multi-format publishing, scientific documents with citations/cross-references, or exporting Jupyter notebooks. Triggers on "render markdown", "render PDF", "publish document", "create presentation", "quarto render", or multi-format publishing needs.
report-infra-work
Report an infrastructure change need in platform-infra (Josh's current primary repo for this category, migrating off the legacy easypost-enterprise-platform-infra / enterprise-platform-bootstrap since 2026-09-24) to the infra-dispatcher session instead of working around it inline or building a PR. Use when a need arises for a new GCP resource, IAM grant, DNS record, GitHub org/repo setting, or Jira config change in platform-infra or either of the two legacy repos, and building it isn't part of the current task.
report-tool-work
Report a bug or request a feature in one of Josh's easypost-sandbox tool/MCP-server repos to the bugfix-dispatcher session instead of working around it inline or building it directly. Use when a tool, CLI, or MCP server under Josh's easypost-sandbox GitHub org is broken, or when it's missing something it should have, and fixing/building it isn't part of the current task.
research
Delegate research and context-gathering tasks to a sub-agent to protect the primary context window. Use when the user asks to "research X", "look into X", "find out about X", "gather context on X", or any investigative framing where answering requires 2+ searches or multiple sources. Also use proactively before starting substantive work when prior context is unknown. Never run research inline — always delegate.
rust
Rust development with cargo - build strategy, clippy triage, testing workflows, and lint configuration. Use when writing, testing, or debugging Rust code and optimizing build efficiency. Prefer cargo check over build and avoid release builds unless explicitly needed. Don't use for C/C++, Go, or other compiled languages.
trim
Trim a prose document (README, design doc, blog post, notes) for readability by cutting redundancy, filler, and dead weight in the author's own words. Use when asked to simplify, condense, or strip a document down to its load-bearing ideas. Not for source code, data files, or summarization.
ui-review
Review rendered web UIs against project layout rules and Josh's design preferences; capture screenshots, repair findings, and turn his feedback into reusable rules and approved references. Use for UI design, layout refinement, or explicit /ui-review requests.
xlsx
Use xlsx binary for Excel file manipulation including viewing, SQL-like filtering, cell editing, conversion to/from CSV, and data analysis operations. Use this instead of Python (openpyxl, pandas, xlrd) or Node.js libraries for ALL .xlsx file operations — reading, writing, filtering, conversion, and analysis. Trigger on any request involving .xlsx or Excel files.
xsv
Fast CSV/TSV processing with xsv - select, search, slice, sort, join, stats, frequency, and indexing for large delimited files. Use when processing CSV-shaped data on the command line with filtering, joining, or summary statistics. Prefer over pandas or awk for CLI-based CSV work. Don't use for Excel formats, multi-sheet data, or complex SQL queries — use xlsx for Excel, duckdb for SQL operations.
bigquery
Use bigquery CLI (instead of `bq`) for all Google BigQuery and GCP data warehouse operations including SQL query execution, data ingestion (streaming insert, bulk load, JSONL/CSV/Parquet), data extraction/export, dataset/table/view management, external tables, schema operations, query templates, cost estimation with dry-run, authentication with gcloud, data pipelines, ETL workflows, and MCP/LSP server integration for AI-assisted querying and editor support. Modern Rust-based replacement for the Python `bq` CLI with faster startup, better cost awareness, and streaming support. Handles both small-scale streaming inserts (<1000 rows) and large-scale bulk loading (>10MB files), with support for Cloud Storage integration.
gspace
Use gspace CLI (command-line tool) and MCP tools for Google Workspace operations including Drive file management, Gmail, Docs, Sheets, Calendar, and Tasks. Both CLI commands (via Bash) and 40+ MCP tools are available.
jira
Use jira CLI for Jira operations including issue management, project queries, transitions, and JQL search
lancer
Use lancer CLI for LanceDB semantic and multi-modal search with document ingestion, vector embeddings, and MCP server integration for knowledge retrieval.
phab
Use phab CLI for Phabricator task management, code review workflows, revision creation and updates, with Git integration and auto-detection features.
qmd
Use qmd for workspace search — hybrid BM25+vector search with Qwen3-4B reranking over indexed workspace documents. Not to be confused with Quarto documents (.qmd files).
az
Use az CLI for Azure cloud resource management, Azure DevOps operations, VMs, storage, networking, AKS, and Key Vault with comprehensive authentication and output control.
claude-cli
Use claude CLI for interactive AI sessions, scripting with print mode, MCP server management, and plugin configuration. Master session management, tool control, and automation workflows.