cracking-shells

A research team for Claude Code: a principal investigator who follows what does not add up, and colleagues who explore, author claims and verify them independently. Claude Code only; macOS and Linux.

3 skills · 5 agents · hooks

Semantic and hybrid code search for coding agents, exposed as MCP tools backed by colgrep.

5 skills · hooks

Maintainer skills for this repository (policy, campaign leadership, landing and release, stack traps). Not for end users.

4 skills

Coins a precise new word and its morphological family for a concept that lacks a compact name.

1 skill

Designs software tests and statistical tests before they are written, so each answers a stated question and can fail.

1 skill

Turns a Claude Design figure or table into a print-quality image placed, undistorted, into a local document.

1 skill

Creates, executes, and amends graph-based campaign roadmaps.

1 skill

Converts digital-native PDFs into Markdown with figures embedded where they belong, using poppler and pandoc.

1 skill

Turns a repository into an installable plugin for Claude Code, Codex, and Agent Plugins 1.0 clients.

1 skill

Writes waits and watches in Claude Code that end by themselves and catch failures, with a sweep that stops stray processes.

1 skill

Produces layered design canons a team uses to make assets in a coherent style and stakeholders use to review them.

1 skill

Guides the git-discipline lifecycle for commit history: vocabulary, versioning, commits, and branch integration.

1 skill

Drafts, revises and reviews prose signed in the author's own voice, with register control against LLM-default rhetoric.

1 skill

Runs the release cycle: integration lanes into main, pull requests as merge commits, CI-cut releases, tooling by repo contents, channels, and recovery.

1 skill

Guides production of stakeholder-reviewable reports for coding, research, and writing work.

1 skill