maindala

Install mAIndala Skills and Agents into your AI coding environment.

Install

npx maindala install <skill-slug>
npx maindala install agent/<agent-slug>

Authentication

Skills and Agents require a personal API key. Get one at https://www.maindala.com/profile.

# Save key permanently
npx maindala login mk_your_api_key_here

# Or use environment variable
MAINDALA_API_KEY=mk_... npx maindala install <slug>

Formats

Auto-detected from your project, or specify with --format:

| Format | Tool | Output | |--------|------|--------| | claude-skill | Claude Code (native skills) | .claude/skills/<slug>/SKILL.md | | claude | Claude Code | CLAUDE.md | | cursor | Cursor | .cursor/rules/<slug>.mdc | | copilot | GitHub Copilot | .github/copilot-instructions.md | | windsurf | Windsurf | .windsurfrules | | openclaw | OpenClaw | .openclaw/skills/<slug>.md | | continue | Continue | .continue/rules/<slug>.md | | cline | Cline / Roo | .clinerules/<slug>.md | | zed | Zed | .rules | | raw | Any | ./<slug>.md |

Table order matches auto-detection priority (top to bottom). This list is checked against the CLI's actual supported formats by scripts/verify-readme-formats.mjs, which fails if this table and the CLI's own Format list ever drift apart again.

Examples

# Install a skill into Claude Code (auto-detected if CLAUDE.md exists)
npx maindala install linkedin-connection-request

# Install an agent into Cursor
npx maindala install agent/stock-analyst-pro --format cursor

# Install with a specific API package format
npx maindala install full-equity-research-report --api-format openai

Discover agents (scan)

# Local report — no account, no API key, works offline
npx maindala scan .

# Machine-readable report, signable before anything touches mAIndala
npx maindala scan . --format sarif --output scan.sarif
# writes scan.sarif + a sidecar manifest.json (SHA-256 digest) — sign the
# digest with your own infrastructure (cosign, GPG, an in-house TSA)

# Optionally anchor the digest with a free public RFC 3161 timestamp
npx maindala scan . --format sarif --output scan.sarif --timestamp

# Push findings into a governed org registry (requires `maindala login <mk_...>`)
npx maindala scan . --register --org my-org

Detects, at three confidence tiers, agent/LLM framework usage in the current directory: dependency manifests (package.json, requirements.txt, pyproject.toml, go.mod), MCP client config / crew-agent definition files (.mcp.json, claude_desktop_config.json, crew.yaml, agents.yaml), and (lowest confidence, never auto-registered) source-code patterns. MCP servers found in a config file are matched against the public mAIndala catalog for context (trust status, check count) — see DATA.md for exactly what is sent over the network in each mode, and to whom.

Releasing

See RELEASING.md — publishing runs through a GitHub Release + trusted-publishing CI workflow with a required-reviewer approval gate, not a local npm publish.