The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to run local Python scripts, such as scripts/rice_prioritizer.py and scripts/customer_interview_analyzer.py. These scripts are referenced in references/agent-registry.md and references/collaboration-templates.md but are not included in the skill package, making their behavior unverifiable.
[DYNAMIC_EXECUTION]: The "Automation Engineering Agent" role, defined in references/opc-agents.md, is explicitly designed to generate and implement executable code. This includes creating Vue 3 frontend components, FastAPI backend routes, and PostgreSQL database schemas. Executing AI-generated code without proper sandboxing or strict human review poses a risk of executing unintended or malicious logic.
[INDIRECT_PROMPT_INJECTION]: The skill demonstrates a significant attack surface for indirect prompt injection by ingesting and processing untrusted data from external sources.
Ingestion points: The agents ingest data from files like interview.txt and features.csv as described in references/agent-registry.md and references/collaboration-templates.md.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat external data as untrusted or to ignore embedded instructions within that data.
Capability inventory: The skill has the capability to execute shell commands and generate executable source code, which could be abused if an injection is successful.
Sanitization: The skill does not define any mechanisms for sanitizing, escaping, or validating the content of external data before processing it.
[CREDENTIALS_UNSAFE]: In references/opc-agents.md, the "Automation Engineering Agent" defines deployment configurations that include environment variables for sensitive credentials, such as OPENAI_API_KEY and DATABASE_URL. While these are intended as placeholders, the instructions encourage the management of secrets in a manner that requires careful handling to avoid exposure.

agent-team