# Blattman Claude Code Skills Review ## Source GitHub: [chrisblattman/claudeblattman](https://github.com/chrisblattman/claudeblattman) Website reference: [Skill Library](https://claudeblattman.com/setup/skill-reference/) ## What it is The repository is a public Claude Code workflow kit aimed at academics and professionals who do not want to build a full system from scratch. It contains: - `skills/` for slash-command style markdown workflows. - `agents/` for specialized Claude Code subagents. - `templates/` for starter config files such as `CLAUDE.md`, goals, email policy, calendar policy, and triage config. - `docs/` for the published site. The README presents three paths: essentials, toolkit, and building custom workflows. ## Useful patterns for this setup - The repo is strongest as a comparative pattern library, not as something to import wholesale. - The useful ideas are mostly architectural: skill libraries grouped by workflow, explicit install snippets, config templates separated from workflow instructions, and reusable review agents. - The `skill-patterns` documentation is worth keeping in mind when reviewing local skills: policy-first architecture, config/state separation, graceful degradation, confirmation gates, and explicit error handling. - The agents are interesting mainly as examples of narrowly-scoped review roles. The `review-writing` agent is the closest to Nick's research workflow, though it would need adaptation to Nick's voice and source-work rules. ## What not to import directly - Do not copy the skills into this setup without adaptation. Nick's vault already has stricter project, source, Obsidian, and epistemic rules. - Do not import executive-assistant workflows that assume Gmail/Calendar/Reminders conventions without reconciling them with the existing `gws`-based commands. - Do not import writing-review workflows without adding Nick's voice files, source-first extraction rules, and the prohibition on treating exploratory philosophy as finished structure. - Do not add another goals or task system unless it clearly replaces something. The local setup already has TaskNotes, Google Tasks sync, Sessions, Dashboard, and project notes. ## Best candidate to adapt The most useful import target is not a finished skill but a review checklist for local skills: - Is the skill's trigger precise? - Does it load the relevant local knowledge file before acting? - Does it separate durable config from transient state? - Does it define what to do when a tool or MCP server is unavailable? - Does it include confirmation gates for irreversible or user-visible actions? - Does it keep critical instructions near the top? ## Recommendation Use the repo as a benchmark when improving this system's skills and agents. Do not install the library as-is. The local setup is more specialized, especially around Obsidian, source quotation, project sessions, and Nick's writing practice.