chore: update skill activation rule

[EdwardIrby]

No description provided.

[claude]

Code Review - PR #25: Update skill activation rule

Summary

This PR enhances the skill activation rule to be more comprehensive and actionable. The changes clarify when and how agents should evaluate and activate skills.

Positive Changes ✓

1. Clearer scope - "Evaluate on every prompt" is more explicit than "Evaluate before implementing"
2. Broader applicability - "Applies to all tasks" removes ambiguity about when the rule applies
3. Better terminology - "Respond" vs "Implement" in step 3 is more accurate for non-code tasks
4. Decision-point re-evaluation - Excellent addition that covers critical moments (plan mode, subagents, task transitions, domain shifts)
5. Improved verification pattern - The verify/fix pattern is now more specific and actionable

Observations & Questions

Scope expansion:
The change from "before implementing" to "on every prompt" is significant. This means:

• Agents should evaluate skills before any response (not just code changes)
• Could apply to simple questions, explanations, research tasks

Question for consideration: Is the intent that agents should evaluate skills even for simple informational queries? For example:

• User: "What version of Bun is required?"
• Agent: Should evaluate all skills before answering?

If yes, the rule is clear. If no, might want to add an exception for trivial informational queries.

Decision-point re-evaluation:
This is a strong addition. The examples are concrete and cover key scenarios where skills might become relevant mid-task.

Minor suggestion: Consider whether "when domain shifts" needs more concrete examples beyond the two provided (code→evaluation, schema→grading). Current examples are clear but might not cover all shift patterns.

Code Quality ✓

• Follows documentation standards from rules/documentation.md
• Uses proper markdown formatting
• Verification patterns are consistent with other rules
• No breaking changes to existing references

Testing Considerations

Since this is a rule for agent behavior (not code):

• No unit tests needed
• Verification would come from observing agent behavior in practice
• Consider monitoring whether agents actually follow the enhanced guidance

Recommendation

Approve with optional consideration: The changes are well-reasoned and improve clarity. The only question is whether "evaluate on every prompt" needs qualification for trivial queries, but this may be intentional to ensure comprehensive skill usage.

The rule now provides clearer guidance that should reduce instances of agents skipping relevant skill activation.