SKILL: Skip Skill Listing Without Skill Tool

Domain: command-surfaces Trigger: Apply when an attachment or context injector is about to advertise capabilities to an agent that may not have the tool needed to use them. Source Pattern: Distilled from reviewed command-surface and listing-control patterns.

Core Method

Before generating capability-discovery context, verify that the receiving agent actually exposes the execution surface for it. If the agent lacks the Skill tool, return no skill-listing attachment at all rather than spending tokens on a catalog it cannot invoke. This keeps the prompt focused on actionable affordances and avoids confusing agents with instructions they have no mechanism to execute.

Key Rules

• Check tool availability before doing any skill-listing work, including command collection or formatting.
• Treat missing execution surfaces as a hard stop, not as a degraded listing mode.
• Keep the guard close to the listing entrypoint so later refactors cannot accidentally compute listings for unsupported agents.
• Prefer actionable context over informative-but-unusable context; if the agent cannot call the tool, omit the attachment.

Example Application

If a background helper agent is configured only with file and shell tools, skip injecting a skill catalog even if the parent session supports skills. The helper sees only the tools it can actually use, while the main agent continues receiving skill-listing updates normally.

Anti-Patterns (What NOT to do)

• Do not send a skill listing just because the overall application supports skills; the specific consumer must expose the Skill tool.
• Do not compute and then discard the listing after expensive formatting work; guard before the heavy path starts.