Install

Claude Code (one-liner via marketplace)

/plugin marketplace add twitu/indiana-codes
/plugin install dig@indiana-codes

That’s it. Restart Claude Code and /dig shows up in any repo with git history.

Other agents (manual)

Clone the repo once. The skill content is the same for every agent — only the install path differs.

git clone https://github.com/twitu/indiana-codes ~/.local/share/indiana-codes
cd ~/.local/share/indiana-codes

Then pick your agent.

Skip this if you used the marketplace install above. Manual symlink path:

mkdir -p ~/.claude/skills ~/.claude/commands
ln -sf ~/.local/share/indiana-codes/dig         ~/.claude/skills/dig
ln -sf ~/.local/share/indiana-codes/commands/dig.md  ~/.claude/commands/dig.md

Restart Claude Code. Type / and you should see /dig.

mkdir -p ~/.config/opencode/skills ~/.config/opencode/commands
ln -sf ~/.local/share/indiana-codes/dig         ~/.config/opencode/skills/dig
ln -sf ~/.local/share/indiana-codes/commands/dig.md  ~/.config/opencode/commands/dig.md

Restart OpenCode. Type / and you should see /dig.

Cursor has no skills directory, so the command file points the agent at the cloned SKILL.md directly:

mkdir -p ~/.cursor/commands
cat > ~/.cursor/commands/dig.md <<EOF
---
description: Code archaeology — chapter-based history walkthrough
---
Follow the skill instructions in $HOME/.local/share/indiana-codes/dig/SKILL.md to perform a codebase excavation pass on this repository.
EOF

Project-scoped install (commits with the repo) is the same, but writes to .cursor/commands/ instead of ~/.cursor/commands/.

Use it

In your agent of choice, in any repo with git history:

/dig

The agent surveys the repo, proposes chapters, and asks which one you want to start with. Then:

Tell me about chapter 1.

Why didn't they just use Postgres here?

Was the v2 API ever finished, or is it abandonware?

Skip ahead to the chapter about the auth migration.

It answers. You push back. By the end of each chapter, you know the chapter — and a markdown file knows it too.

In a fresh chat, no need to re-run /dig — just point the agent at the existing PROJECT_HISTORY.md:

Based on PROJECT_HISTORY.md, walk me through chapter 3.

Connect more sources

Code Archaeology uses whatever it can find. More sources = better chapters.

Source	How
Git history	Built in. Always on.
GitHub issues / PRs / discussions	Local `gh` CLI works out of the box. The official GitHub MCP is even better.
GitHub releases	Same: `gh release` or the MCP.
Jira / Linear / Confluence / Notion	If your team has the corresponding MCP configured in your agent, the skill picks it up automatically.
Local docs	`docs/`, `adr/`, `rfc/`, `CHANGELOG.md` — read automatically.