11 replies

Per-Agent Skill allowlist

✅Solved

I wasn't able to find an example showing how to configure skills on a per-agent basis. I have many skills defined globally, but for a particular agent, I'd like to limit it to a single skill.

Solution

        id: "nc-only",
        // Only these skills will be eligible / injected for this agent
        skills: ["nextcloud"],
      },
    ],
  },
}

        id: "nc-only",
        // Only these skills will be eligible / injected for this agent
        skills: ["nextcloud"],
      },
    ],
  },
}

That matches what you’re observing in

/context

/context

(only

nextcloud

nextcloud

injected into the system prompt).

Notes / gotchas

- The skill still has to exist in one of the normal skill locations (bundled,

~/.openclaw/skills

~/.openclaw/skills

, or that agent’s

<workspace>/skills

<workspace>/skills

), and pass any gating (

requires.*

requires.*

etc.).
- If you also use per-channel

skills: [...]

skills: [...]

filters (Discord/Slack/Telegram support this), that will further restrict what’s exposed in that chat (so you can do agent-wide + chat-specific narrowing if you want).

If this solved it for you, please right-click the solution message → Apps → Mark Solution so Answer Overflow indexes it.

Jump to solution

FotC�

Friends of the Crustacean 🦞🤝•2w ago•

11 replies

Norbert

Per-Agent Skill allowlist

✅Solved

I wasn't able to find an example showing how to configure skills on a per-agent basis. I have many skills defined globally, but for a particular agent, I'd like to limit it to a single skill.

Solution

        id: "nc-only",
        // Only these skills will be eligible / injected for this agent
        skills: ["nextcloud"],
      },
    ],
  },
}

        id: "nc-only",
        // Only these skills will be eligible / injected for this agent
        skills: ["nextcloud"],
      },
    ],
  },
}

That matches what you’re observing in

/context

/context

(only

nextcloud

nextcloud

injected into the system prompt).

Notes / gotchas

- The skill still has to exist in one of the normal skill locations (bundled,

~/.openclaw/skills

~/.openclaw/skills

, or that agent’s

<workspace>/skills

<workspace>/skills

), and pass any gating (

requires.*

requires.*

etc.).
- If you also use per-channel

skills: [...]

skills: [...]

Jump to solution

Per-Agent Skill allowlist

Notes / gotchas

Similar Threads

Per-Agent Skill allowlist

Notes / gotchas

Similar Threads

Similar Threads

Similar Threads