How this works: every tool defaults to Allow and runs without a prompt.
Flip a tool to Ask to require a one-time approval each run, or Block to deny it.
These are enforced live in agent & War Room runs.
No tools manifested.
Upload a skill
Upload a Claude Code SKILL.md file (or paste its contents). The frontmatter (name + description) is read automatically. Once uploaded, any agent can invoke this skill via step dispatch.
No file selected — or paste below
Skill
This is the full SKILL.md playbook the agent/team follows when this skill runs.
Loading…
Generate a skill
Describe what the skill should do in plain English. We'll draft a SKILL.md you can edit before saving.