agent-skills

Check whether a known credential has already leaked publicly via GitGuardian's HasMySecretLeaked (HMSL) hash-lookup service. Use when the user inherits credentials, suspects a specific token leaked, wants to vet a HashiCorp Vault inventory, or asks "has this secret leaked", "is this compromised", or "check against HMSL". Distinct from scan-secrets, which finds unknown secrets in code; this checks known secrets against the HMSL corpus. User-run only — this is a command-handoff skill, so the agent prepares the commands, the user runs them in their own terminal, and the agent only interprets the sanitized output the user pastes back.

Use when generating or planting GitGuardian honeytokens, canary tokens, decoys, or tripwire credentials. Use around .env.example, sample configs, pre-publication open-source repos, internal wikis, runbooks, deploy scripts, or other attractive leak surfaces.

Install ggshield as a hook so secrets are caught before they leak. Covers git hooks (pre-commit and pre-push) that block secrets from entering git history, and AI-assistant hooks (claude-code, codex, copilot, cursor, vscode) that scan an AI coding tool's prompts, actions, and outputs in real time. Use when asked to install Claude Code, Cursor, Copilot, or git hooks, or to block secrets from being committed or pushed.

Scan a developer's entire machine for credentials across local git repositories, dotfiles, ~/.aws/credentials, ~/.kube/config, ~/.docker/config.json, ~/.npmrc, browser profile databases, shell history, AI agent caches, and abandoned project trees via ggshield machine scan. Use when preparing to wipe, sell, or hand off a laptop, when onboarding a new machine, when auditing what credentials live on a machine, or when the user explicitly asks to inventory secrets on the system itself. Requires endpoint scanning enabled on the GitGuardian workspace; not available on Free.

Use when scanning code, commits, git history, Docker images, or packages for hardcoded secrets, when editing credential-handling code, .env files, CI/CD workflows, Dockerfiles, or deployment scripts, or before committing or pushing.

Use when triaging or reviewing GitGuardian secret incidents already detected in the dashboard, when asked what is leaking in the org or what to fix first, when remediating or rotating a credential flagged in an incident, after a Public Monitoring alert, or to assign, tag, or resolve incidents. Operates through the GitGuardian Developer MCP server.