ANMA
github.com/anma-labs/anma →Plain-YAML contracts that keep your AI coding agent on a leash
WHAT IT SOLVES
AI coding agents are getting scary-powerful — but would you trust one not to touch prod configs or rewrite your deploy script? Most people draw mental boundaries. This writes them down
WHY IT'S INTERESTING
One YAML file, three enforcement points
You write one YAML boundary declaration. ANMA compiles it into three things: a CLAUDE.md instruction file the agent actually reads, git hooks that block violations, and CI checks that catch them in PRs. It's not another linter — it's moving 'what the AI shouldn't touch' out of your head and into the repo
132 commits and 4 active branches
This isn't a weekend demo. There's a dogfood CI action, a test harness, hotfixes for edge cases like exclude-pattern honoring. The HN title uses 'boundary contracts' — the author is clearly trying to define a category, not ship a throwaway script
「「Boundary contracts for cheaper AI coding agents」— from the Show HN title」
TECH GUESS
Likely Python or Node, outputs Markdown hooks and CI YAML; probably uses AST parsing to generate the CLAUDE.md boundary file
Discussion (0)
- No comments yet — be the first.
Related
#229▶ 136Upload your textbook. AI teaches it back to you, chapter by chapter
#228▶ 112Someone needs to watch the AI spend meter
#227▶ 184Your users are kittens. Don't let them fall off the cliff