feat(clink): Allow sub-agents to inherit project directives and context #384
Description
What problem is this feature trying to solve?
When sub-agents are spawned using [clink], they currently operate in isolation from the project's specific configuration files and directives. This means the LLM (e.g., Claude or Codex) cannot access or "see" the pre-defined rulesets (like AGENTS.md or CLAUDE.md ) or custom agent instructions already established in the project root. Consequently, the sub-agents fail to adhere to project-specific standards and architectural constraints, leading to less effective and inconsistent outputs that require manual correction.
Describe the solution you'd like
I would like to request a mechanism that allows sub-agents to be executed directly within the project's run scope. The solution should ensure that any spawned sub-agent automatically inherits and accesses the project's full context, including:
Pre-defined rulesets (e.g., AGENTS.md or CLAUDE.md).
Custom agent directives and operational guidelines defined in the project.
By sharing this context, the agents can leverage the existing instructions to generate code that is immediately compliant with the project's standards, making the workflow significantly more efficient.
Describe alternatives you've considered
No response
Feature Category
Workflow improvement
Contribution
- I am willing to submit a Pull Request to implement this feature.