[Feature]: Detect conflicts across project context and skills

Problem or Motivation

As projects use more sources of agent context—such as AGENTS.md, project-specific instructions, enabled skills, and external skills—the chance of conflicting or inconsistent guidance increases.

Today, users may enable a skill that contradicts existing project instructions without any warning. For example, AGENTS.md may require one architectural pattern while an enabled skill recommends another.

These conflicts can lead to inconsistent agent behavior, incorrect implementation decisions, and unnecessary rework.

Desired Behavior

OpenHands should be able to analyze the active project context and surface potential conflicts between sources such as AGENTS.md, enabled skills, project instructions, and repository conventions.

The system should show the conflicting instructions, explain why they may be inconsistent, and ask the user to choose how the conflict should be resolved.

It should also be able to flag likely mismatches between written instructions and the actual repository, such as an instruction referencing an architecture, framework, or pattern that the project no longer uses.

Acceptance Criteria

• OpenHands can inspect active context sources such as AGENTS.md, enabled skills, and project instructions.
• Potential semantic conflicts between active instructions are detected and surfaced before execution.
• Each conflict shows the relevant sources and the conflicting guidance.
• Users can choose which instruction should take precedence or disable the conflicting skill.
• Likely mismatches between written instructions and the actual repository can be flagged.
• Conflict detection is advisory and does not block execution unless the user chooses to resolve it first.
• When no meaningful conflict is found, the normal workflow remains unchanged.

Alternatives Considered

No response

Additional Context

No response