aiignore-cli — Summary

aiignore-cli is a TypeScript/Node.js CLI tool (binary: aiignore, npm package: aiignore-cli) that auto-detects which AI coding tools are active in a project and generates the correct ignore-file configuration for each one — .cursorignore, .geminiignore, .aiignore, .codeiumignore, .aiderignore, .clineignore, .rooignore, and Claude Code's settings.json deny rules — all from a single command. The tool's real value is the embedded security research: documented CVEs, bypass vectors, and per-tool reliability ratings for each AI tool's ignore mechanism, so developers don't need to research tool-specific quirks individually. It supports a verify command that produces a protection-status table with CI-mode exit codes, enabling ignore-coverage checks in pre-commit hooks or CI pipelines. This is a developer-facing security scaffolding tool that works across 9 AI coding tools.

Differs from seeds: No seed is architecturally similar — aiignore-cli is a project-scaffolding CLI for secrets protection, not a development workflow framework. The closest distribution type is agent-os (bash-bundle, project configuration generator), but aiignore-cli is a compiled TypeScript npm package with a real binary. Unlike all seeds which configure AI agent behavior, aiignore-cli configures AI agent access restrictions.

aiignore-cli — Origin, Philosophy, and Manifesto

Origin

Built by yjcho9317. TypeScript/Node.js, Apache-2.0 license. Single contributor. Version 1.2.0.

Philosophy

The core insight: each AI coding tool has a different, incompatible ignore mechanism with different reliability characteristics and known security vulnerabilities. Developers who rely on a single .gitignore-style file are unknowingly exposing secrets to most AI tools they use.

From README:

"Do you know that Cursor also needs .cursorignore, Claude Code needs settings.json deny rules, Gemini CLI needs .geminiignore, JetBrains needs .aiignore, and Windsurf still uses .codeiumignore? Do you know that Cursor's ignore is 'best-effort' with 2 known CVEs, that Gemini's negation patterns are broken, or that Copilot has no ignore file at all? aiignore does the research for you. The security data behind each tool is the real value — the CLI just applies it."

Security-first positioning

aiignore treats AI tool ignore mechanisms as a security surface, not just a UX convenience. It documents known bypass vulnerabilities explicitly:

• Cursor: "best-effort", agent bypass, @ reference bypass
• Gemini CLI: negation patterns broken, self-blocks .env/.pem
• Copilot: no ignore file exists for individual developers at all

Limitation acknowledgment

From README: "No AI tool guarantees 100% file exclusion. All tools share a common weakness: agent/terminal modes can bypass ignore files by running shell commands directly. Copilot has no ignore mechanism at all for individual developers. This tool is one layer of defense."

This honest limitation statement positions aiignore as one defense layer, recommending complementary tools (gitleaks, trufflehog, secrets managers).

aiignore-cli — Architecture, Distribution, and Installation

Distribution

npm package: aiignore-cli. Binary: aiignore (and alias aiignore-cli).

npx aiignore-cli init
npm install -g aiignore-cli

Version analyzed: 1.2.0

Directory Tree

aiignore-cli/
├── src/
│   ├── index.ts            # entry point
│   ├── commands/           # init, verify, list, config commands
│   ├── generators/         # per-tool config file generators
│   ├── scanners/           # tool detection scanners
│   ├── data/               # security research data per tool
│   └── utils/
├── test/                   # vitest test suite
├── docs/
│   └── test-report.md      # per-tool CVEs, bypass bugs, tested behavior
├── assets/                 # screenshots for README
├── package.json            # bin: { aiignore, aiignore-cli }
├── tsconfig.json
└── tsup.config.ts          # build config

Tool Support (9 AI tools)

Required Runtime

• Node.js >= 18

Config Files

• .aiignorerc — per-project configuration (tools list, extraPatterns)
• ~/.config/aiignore/config.json — global configuration (personal patterns)

Patterns Source

Built-in defaults + security-related entries from .gitignore. Categories: env files, credentials, keys, SSH, cloud provider configs, infrastructure secrets, app secrets, database files, certificates.

aiignore-cli — Components

CLI Commands

aiignore init (default)

Auto-detect AI tools in the project and generate appropriate ignore files.

Flags:

• --all — generate for all supported tools, skip detection
• --only cursor,gemini — specific tools (comma-separated)
• --append — add missing patterns to existing files
• --dry-run — preview without writing
• --force — overwrite existing files
• -q — quiet mode

aiignore verify

Produce a protection-status table showing which tools are protected.

Flags:

• --ci — exit 1 if any unprotected tool detected
• --strict — exit 1 if any tool isn't best-effort or better
• --json — machine-readable output

aiignore list

Show all supported tools and their aliases.

aiignore config

Show effective configuration (merged global + project).

Flags:

• path — print global config file path

Generators (per tool)

One generator module per supported AI tool in src/generators/. Each generator knows:

• Which file to write
• Which pattern format the tool uses
• Known quirks (e.g., Gemini negation broken, Cursor best-effort caveat)

Scanners

src/scanners/ — auto-detect which AI tools are present in the project (checks for tool-specific config files, lock files, etc.)

Security Data

src/data/ + docs/test-report.md — per-tool security research: CVEs, known bypass vulnerabilities, tested behavior. This is the embedded knowledge base that differentiates aiignore from a manual script.

Project Config: .aiignorerc

{
  "tools": ["cursor", "claude", "jetbrains"],
  "extraPatterns": ["internal/", "*.staging.env"]
}

Global Config: ~/.config/aiignore/config.json

{
  "extraPatterns": ["company-internal/", "*.corp-secret"]
}

aiignore-cli — Uniqueness and Positioning

Differs from Seeds

No seed is architecturally similar — aiignore-cli is a developer-facing security configuration scaffolding tool, not a development workflow framework. It operates on the meta-level: configuring what AI tools are allowed to see, not how AI agents behave. The closest seed by distribution type is agent-os (generates project configuration files from a CLI), but agent-os generates CLAUDE.md context files while aiignore-cli generates security restriction files. No seed mentions cross-tool ignore file management. Unique in corpus for being the only tool that treats different AI tools' file-access restrictions as a security surface to be systematically protected.

Positioning

aiignore-cli sits at the intersection of developer security hygiene and AI tool adoption. It solves a real practical problem: the proliferation of AI coding tools has created a parallel proliferation of incompatible ignore-file mechanisms, each with different reliability characteristics and undocumented bypass vulnerabilities. A developer using three AI tools (e.g., Claude Code, Cursor, Cline) needs three different configuration files in three different formats with different quirk knowledge.

The tool's competitive moat is the embedded security research in docs/test-report.md — the reliability ratings and documented bypass vectors are based on actual testing, not vendor documentation.

Observable Failure Modes

1. Single maintainer, low star count (8): This is a niche tool with minimal community validation. Security claims may be outdated.
2. No bypass guarantee: README explicitly states terminal commands can always bypass ignore files by running shell commands directly. This is an inherent limitation of the approach.
3. Copilot blind spot: GitHub Copilot has no ignore mechanism for individual developers — aiignore can only generate a "guide only" for Copilot users, not a functional protection file.
4. Static pattern list: Patterns are static (baked into the package). Project-specific sensitive files not in the default list require manual configuration via .aiignorerc.
5. Gemini CLI negation bug: The README documents that Gemini's negation patterns are broken. aiignore works around this by not generating negation patterns for Gemini, but the underlying tool limitation persists.

Explicit Antipatterns

• Treating aiignore as the sole layer of secrets protection (README explicitly recommends gitleaks, trufflehog, and secrets managers as complementary layers)
• Assuming 100% file exclusion from any AI tool