The Agent Skills Directory

[COMMAND_EXECUTION]: The skill integrates the mcp__rube__RUBE_REMOTE_BASH_TOOL, which allows the agent to execute arbitrary shell commands on a remote system as part of automated workflows.
[REMOTE_CODE_EXECUTION]: The mcp__rube__RUBE_REMOTE_WORKBENCH tool provides a Python sandbox environment, enabling the execution of dynamic Python scripts at runtime to process data.
[COMMAND_EXECUTION]: The mcp__pal__clink tool facilitates external CLI integration, extending the agent's ability to run system-level commands through its reasoning component.
[REMOTE_CODE_EXECUTION]: The skill includes persistence mechanisms via mcp__rube__RUBE_MANAGE_RECIPE_SCHEDULE, allowing the creation and scheduling of automated workflows using cron expressions that can execute arbitrary code over time.
[PROMPT_INJECTION]: The skill presents a high risk for indirect prompt injection due to its complex integration surface:
Ingestion points: Data enters the agent context from over 500 integrated applications (e.g., Slack messages, GitHub PR bodies, Jira tickets) as described in SKILL.md.
Boundary markers: The skill definition lacks explicit delimiters or instructions to ignore embedded commands within external data payloads.
Capability inventory: The agent has access to highly privileged tools including REMOTE_BASH_TOOL and REMOTE_WORKBENCH which can be triggered by these workflows.
Sanitization: There is no evidence of sanitization or validation of the content retrieved from third-party services before it is used to influence tool parameters or execution logic.

sc-mcp