Pristine — AI skill for Claude Code
PR Scope & Intent Guardrail for Fast-Moving Teams & AI Agents.
How to install Pristine
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open ruslanlap/pristine and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Pristine does
PR Scope & Intent Guardrail for Fast-Moving Teams & AI Agents. A tiny, zero-dependency, read-only Rust CLI (< 5ms cold start, zero network) that catches non-matching files sneaking into docs:, ci:, test:, or style: pull requests and commits.
Alternatives in AI
README
pristine
 [](https://crates.io/crates/pristine-scope) [](LICENSE)
**PR Scope & Intent Guardrail for Fast-Moving Teams & AI Agents.** A tiny, zero-dependency, read-only Rust CLI (`< 5ms` cold start, zero network) that catches non-matching files sneaking into `docs:`, `ci:`, `test:`, or `style:` pull requests and commits.
┌──────────────────────────────────────────────────────────────────────────┐
│ $ pristine check --base main --title "docs: fix typo in README" --agent │
│ │
│ [PRISTINE VIOLATION] Branch/PR intent is 'docs:', but non-documentation │
│ path(s) [src/auth.rs, .github/workflows/release.yml] were modified. │
│ Action required: Revert [src/auth.rs, .github/workflows/release.yml] │
│ or change the PR/commit prefix to 'feat:' / 'fix:' / 'chore:'. │
└──────────────────────────────────────────────────────────────────────────┘
Why `pristine`?
Modern engineering teams and open-source maintainers face three growing risks:
- Trojan PRs & AI Slop (
xz-style supply chain & agent drift): A PR titleddocs: fix typo in READMEmodifies 20 markdown files—and quietly changes one line in.github/workflows/release.yml,Cargo.toml, orsrc/auth.rs, or plants a symlink insidedocs/pointing to sensitive code. - Broken CI Fast-Tracks: Fast-moving teams often configure auto-merge or skip heavy 40-minute E2E test suites on
docs:pull requests. If a developer or an AI coding agent (Claude Code, Cursor, Codex, Devin) accidentally touches application code in adocs:PR, broken code ships straight to production untested. - Corrupted Semantic Releases: When source code sneaks into a
docs:,ci:, ortest:PR,release-please/ `sema
Related Skills
AI Outreach Automation
Claude Code skill that drafts governed, CASL-compliant cold outreach at scale: reusable prompt library, guardr
Beyond The Chatbox
A field guide for moving beyond web chat AI — Claude Desktop and Claude Code setup, context engineering, and p
Usewarden
Free, local guardrail for AI coding agents — Claude Code, Cursor, Gemini CLI and more. Blocks out-of-scope wri
Skillgrab
One command. Scan any project and auto-install matching AI agent skills from skills.sh. Zero config. Detects J
Datafog Python
Offline PII firewall for AI agents and LLM apps: fast local detection and redaction, Claude Code hook, LiteLLM
DblCheck
AI-assisted network intent validation framework for multi-vendor environments. Continuously checks live networ
Related Agents
Cold Reviewer
Zero-context code reviewer. Reviews changes cold - no spec, no project knowledge, no domain context. Catches w
Cold Verifier
Phase 11 cold verification agent that independently re-verifies CRITICAL and HIGH findings with zero prior con
Fpm Industry Vocab
Manufacturing ERP domain expert. Validates technical terms in context — catches "SMS" that should be "sMES", "