Scan code for secrets, injection, SSRF, unsafe crypto, and OWASP Top 10 issues.
Scan code for secrets, injection, SSRF, unsafe crypto, and OWASP Top 10 issues.
After writing code that handles input, auth, or sensitive data.
An agent skill is a packaged instruction set (a SKILL.md + assets) that a model loads on demand when a task matches — so the agent gains a capability without bloating every prompt. Skills compose: a single agent can carry many, and invoke the right one per task.
A skill tells your agent how to do something. To let it act on the world, pair skills with tools: trust-graded MCP servers (we grade every one A–F so you know it's safe to connect) and any website turned into agent tools via WebMCP — no server to build.