The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection surface. The skill functions as an orchestrator, processing outputs from other agent instances to manage complex task lifecycles.\n
Ingestion points: The coordinator agent reads sub-agent status reports via the follow-up API and monitors execution quality using the logs/filter endpoint (documented in references/orchestration.md and references/quality-review.md).\n
Boundary markers: While the instructions use Markdown sections (e.g., ## Requirements) to organize data, they do not include specific delimiters or instructions to the LLM to ignore potentially malicious commands embedded in the sub-agent logs.\n
Capability inventory: The skill has significant agency, including the ability to merge git branches, create recurring cron jobs, and trigger new issue executions across multiple projects (SKILL.md, references/rest-api.md).\n
Sanitization: There is no mention of sanitizing or escaping the data retrieved from the server before it is interpolated into the coordinator's context.\n- [COMMAND_EXECUTION]: Execution of project-defined build and test scripts.\n
Evidence: The post-merge verification workflow in references/merge-strategy.md explicitly calls npm run build and npm run test to validate changes.\n- [SAFE]: Persistence mechanisms via the BKD server's cron API are provided for legitimate operational needs, such as handling long-running tasks and status check-ins (references/rest-api.md).

bkd