The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data via the $ARGUMENTS variable, which includes user-defined phase numbers and descriptions. This input is processed by a workflow that has access to sensitive tools like Write and Bash. This represents a standard attack surface where malicious input could potentially attempt to influence the logic of the roadmap update or the command-line execution.
[COMMAND_EXECUTION]: The skill explicitly requests the Bash tool and directs the agent to execute a workflow from an internal platform path. While the behavior appears consistent with its stated purpose of managing a software roadmap, the use of shell tools to process external input is a capability that warrants least-privilege configuration.

gsd-insert-phase