The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute commands that interpolate variables derived from external scenario files (e.g., $SCENARIO_PROMPT, $SCENARIO_ID, $SCENARIO_NAME). This occurs in Phase 2 and Phase 4. If a scenario file contains shell metacharacters such as semicolons, backticks, or command substitutions, it could lead to arbitrary command execution on the host machine.
[PROMPT_INJECTION]: The skill contains instructions that could be leveraged for indirect prompt injection. It ingests untrusted data from scenario JSON files and uses it as a prompt for a downstream agent dispatch via run_agent.py.
Ingestion points: Test scenario files located in temp/os-evolution-verifier/scenarios/*.json are read and parsed.
Boundary markers: Absent. The data from the JSON fields is directly interpolated into the execution logic and agent prompts without delimiters or warnings.
Capability inventory: The skill has access to the Bash, Read, and Write tools, allowing it to execute scripts, modify files, and perform shell operations.
Sanitization: While the skill validates the JSON structure (e.g., required fields), it does not perform any escaping, validation, or filtering of the content within those fields before they are used in command lines or prompts.

os-evolution-verifier