skill-role-inject-context

To inject minimal context for per-agent payloads, automate the extraction of tailored context packages from the .agent/ directory. This ensures each agent type, like CoderAgent or ReviewerAgent, receives only the specific, relevant data needed for its task.

The token limit for structured YAML context packages is 500 tokens per agent. Enforcing this limit prevents over-injection, guarantees efficiency, and ensures each worker agent receives a compact, clearly structured input payload.

To prevent information leakage when preparing inputs for CoderAgent and ReviewerAgent, enforce task-specific context inclusion and guardrails. Trimming irrelevant data and applying structured YAML frontmatter limits exposure to only necessary fields.

Yes, you can use YAML frontmatter metadata to tailor context for TesterAgent and DocumenterAgent. The orchestrator extracts and packages relevant data using this structured format, satisfying requirements for specific field enforcement and data trimming.

The best way to automate per-agent input payload generation is initializing an orchestrator that extracts relevant context from the .agent/ directory. It applies guardrails and trims irrelevant data to create compact, role-specific packages.

You should not use automated context injection when tasks lack clear role separation or when required context exceeds the 500-token limit. The approach relies on strict YAML frontmatter fields and task-specific guardrails to prevent over-injection.