Please report security issues privately by emailing security@logic.inc.

Do not open a public GitHub issue for a suspected vulnerability. Include as much of the following as you can:

• The skill (directory name) and commit involved.
• The agent coder, OS, and crew version where you observed the issue.
• The command, prompt, or workflow that exposes the issue.
• Reproduction steps, proof of concept, or relevant logs.
• Impact: what an attacker could read, write, execute, or cause an agent or user to trust.

We will review the report, follow up if we need more detail, and coordinate a fix and disclosure timeline.

crew-skills is the default core tap shipped with crew. Every skill here ships verbatim to every user who installs it, and the instructions are read directly by agent coders. Security reports are especially useful for:

• A skill that instructs an agent to execute untrusted shell commands, follow symlinks out of the working tree, exfiltrate environment variables, or write outside the project.
• A skill that contains secrets, private data, internal hostnames, or credentials in examples, fixtures, or reference files.
• Prompt-injection or jailbreak payloads embedded in skill content.
• A skill name, description, or frontmatter shape designed to impersonate or shadow another skill in crew search / crew install.
• Supply-chain issues in the GitHub Actions workflows under .github/workflows/ (e.g. an unpinned action, an injection sink in a workflow input).

Vulnerabilities in the crew CLI itself (installer, tap resolution, update flow) belong in the crew repo's security policy.

Normal bugs, feature requests, copy fixes, and support questions can use GitHub issues.

This is a rolling content repository. Fixes land on main and are picked up the next time a user runs crew search or crew install. There are no versioned branches to back-port to.