[P2-a] Add context: fork + agent to analysis/research skills

[JeremyDev87]

Parent: #731

Purpose

Add context: fork and agent to analysis/research-oriented skills so they run in an isolated subagent. This prevents polluting the main conversation context and returns only independent analysis results.

v2.0 Rule

context: fork: Runs in an isolated subagent. The skill content becomes the subagent's prompt.
agent: Specifies the subagent type. Options: Explore (read-only), Plan (planning), general-purpose (full tools).

Important Note

context: fork only makes sense for skills with explicit instructions. If your skill contains guidelines without a task, the subagent receives the guidelines but no actionable prompt.

Therefore, each skill must be verified to contain explicit task instructions before applying fork.

Target Files (4 files)

1. packages/rules/.ai-rules/skills/code-explanation/SKILL.md

context: fork
agent: Explore

Reason: Code explanation is read-only exploration; no conversation context needed. Explore agent is optimal.

2. packages/rules/.ai-rules/skills/security-audit/SKILL.md

context: fork
agent: general-purpose

Reason: Security audit can run as independent analysis. Requires grep/bash access, so general-purpose.

3. packages/rules/.ai-rules/skills/pr-review/SKILL.md

context: fork
agent: Explore

Reason: PR diff analysis is read-only. Only results are returned to the main context.

4. packages/rules/.ai-rules/skills/performance-optimization/SKILL.md

context: fork
agent: general-purpose

Reason: Profiling analysis is independent but may need bash execution.

Pre-verification Checklist

Verify each skill's body is suitable for context: fork:

• ✅ Does it have explicit task instructions? (step-by-step procedures, analysis processes, etc.)
• ✅ Does it receive analysis targets via $ARGUMENTS?
• ❌ If it only contains guidelines, fork is unsuitable → add task structure to the skill body in this case

How to Apply

1. Verify each skill's body for fork suitability
2. If suitable, add context: fork + agent to the YAML frontmatter
3. If unsuitable, add minimal task structure to the body before modifying frontmatter

Acceptance Criteria

• All 4 files have context: fork + agent added
• Each skill returns meaningful results when run in fork mode
• No YAML frontmatter syntax errors
• No changes to existing fields