#2977: Request: Allow users to manually attach/force a skill before a conversation turn

Problem

Hermes skill triggering is entirely prompt-driven with no code-level enforcement. The model decides whether to load a skill based on the system prompt's <available_skills> list, but this decision is voluntary — the model can ignore relevant skills and answer directly.

This means most skills effectively become dead weight. A user may have a well-written skill for a specific task, but the model will only use it if it voluntarily decides to call skill_view.

Requested Feature

Allow users to manually attach/force a specific skill before or during a conversation turn, so the model is forced to follow that skill's workflow regardless of its own judgment.

Concretely:

1. Add a UI control (dropdown, command, or sidebar panel) listing all available skills
2. When a skill is selected, inject its content into the context before the next model call
3. The model should be explicitly told: "You MUST follow the workflow in this skill"

This ensures skill author intent is respected and removes the dependency on model goodwill.

Why This Matters

• Skill authors spend significant effort writing detailed workflows
• Without enforcement, those workflows are effectively optional
• Users have no way to hold the model accountable to a specific process
• This is especially important for regulated workflows (code review, debugging, planning)

Suggested UX

Option A: /use <skill-name> command — prepends skill to next turn Option B: Sidebar skill picker — click to attach, click again to detach Option C: Both