The Agent Skills Directory

[COMMAND_EXECUTION] (HIGH): The skill includes the run_command tool in its definition. Although the provided operational sequences focus on architecture documentation (ADRs and C4 models), the availability of this tool allows the agent to execute arbitrary shell commands on the underlying system.
[PROMPT_INJECTION] (HIGH): This skill is highly susceptible to Indirect Prompt Injection (Category 8). It is designed to ingest and interpret external business logic, domain models, and feature requirements provided by users or read from files.
Ingestion points: Reads project requirements and existing codebases via read_file and user input.
Boundary markers: None identified. There are no instructions for the agent to ignore or delimit embedded commands within the data it processes.
Capability inventory: Access to run_command, write_file, and edit_file allows for full system compromise if an attacker embeds instructions in a 'System Design' request.
Sanitization: No evidence of input sanitization or validation of external content before processing.
[DATA_EXFILTRATION] (MEDIUM): The combination of read_file (to access sensitive architecture/config files) and the capability to use run_command (which could trigger curl or wget) creates a significant risk for data exfiltration of the host environment or the project's intellectual property.

resonance-architect