Edith Augment Code

Use Augment Code (Auggie CLI) to analyze, generate, and modify code through Edith smart glasses or OpenClaw. Triggers when the user asks to build, code, debu...

Security Scan

The skill's instructions match its stated purpose (wrapping the Auggie CLI) but it omits important details about credential storage and remote behavior and gives the agent broad discretion to run commands that could modify or transmit code.

Purpose & Capability

Name/description align with the runtime instructions: the SKILL.md tells the agent to call the Augment Code CLI (auggie) to generate, analyze, debug, and modify code. The required actions (installing the auggie npm package and running auggie login) are coherent with that purpose and no unrelated binaries or credentials are requested.

Instruction Scope

Instructions explicitly run auggie commands against the 'current project' and specific source paths (e.g., src/ws-relay.ts). Examples use --print (read/analysis), which is safer, but the skill also describes generating and modifying code without clearly limiting commands to read-only modes. The SKILL.md does not disclose whether auggie will send project contents to a remote service, so the agent could transmit sensitive code or modify files if non-print modes are used.

Install Mechanism

This is an instruction-only skill with no install spec. It asks users to run 'npm install -g @augmentcode/auggie' themselves. That is a common install path (npm registry) but is performed outside the skill; the skill does not auto-download or execute remote archives.

Credentials

The skill declares no required env vars or config paths, which is appropriate. However, it requires 'auggie login' (authentication) but does not describe where credentials/tokens are stored or whether the CLI will call remote services. Absence of declared credential handling is a gap: the CLI likely stores tokens locally and/or sends code to Augment Code's backend, which has privacy implications not documented here.

Persistence & Privilege

always:false and user-invocable:true (default) — the skill does not request permanent inclusion or elevated platform privileges. It does not modify other skills or system agent configs according to the provided files.

What to consider before installing

This skill is coherent with its stated purpose (wrapping the Auggie CLI) but exercise caution before using it on private or proprietary code. Key considerations: - Understand what 'auggie login' does: verify where the CLI stores tokens (local files) and how to revoke them. - Verify whether the Auggie CLI sends project files to a remote service (read the CLI's docs and the npm package page) before running analysis on sensitive repositories. - Prefer the --print/read-only mode for analysis and debugging tasks; avoid running commands that may modify files without an explicit confirmation step. - When installing, confirm the npm package identity and maintainers (@augmentcode/auggie) on the official registry; consider installing in an isolated environment if you are unsure. - The skill's source is unknown and there is no homepage link; if you require strong assurance, request the publisher to provide a homepage, privacy/security docs, and details on auth/token storage and network behavior before installing.

Augment Code via Auggie CLI

Use Augment Code's powerful context engine and agentic coding capabilities through OpenClaw. This skill wraps the auggie CLI to let users generate code, analyze projects, debug issues, and build prototypes — hands-free through Edith glasses or any OpenClaw channel.

Prerequisites

• Auggie CLI installed: npm install -g @augmentcode/auggie
• Logged in: auggie login

When to use

Trigger this skill when the user asks to:

• Build, create, or scaffold something ("build me a landing page", "create a REST API")
• Analyze or review code ("analyze this codebase", "review this function")
• Debug or fix issues ("why is this failing", "fix the auth bug")
• Prototype or generate code ("prototype a todo app", "generate a React component")
• Get code suggestions or insights ("how should I structure this", "what's wrong with this approach")

How to use

Generate code or build something

Run auggie in print mode for one-shot tasks:

auggie --print "Build a simple Express API with /health and /users endpoints" 2>&1

Analyze the current project

auggie --print "Analyze this codebase and give me a summary of the architecture" 2>&1

Debug an issue

auggie --print "Look at the recent error in src/ws-relay.ts and suggest a fix" 2>&1

Get quiet/structured output

For clean responses without step details:

auggie --print --quiet "What dependencies does this project use and are any outdated?" 2>&1

Response formatting

When returning results to the user (especially through Edith glasses voice output):

• Summarize the key output concisely
• For code generation: describe what was created and where
• For analysis: give the top 3-5 findings
• For debugging: state the root cause and fix
• Keep it conversational and brief — this may be spoken aloud through glasses

Error handling

• If auggie is not installed: tell the user to run npm install -g @augmentcode/auggie
• If not logged in: tell the user to run auggie login
• If the command times out: suggest breaking the task into smaller pieces
• Auggie may take 10-30 seconds for complex tasks — warn the user if it's a big request