The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied context via the $ARGUMENTS variable and reads existing content from docs/roadmap.md. While this constitutes a surface for indirect prompt injection, the skill lacks sensitive capabilities such as network access, shell command execution, or privileged file system operations that could be exploited by malicious input.
[DATA_EXPOSURE]: The skill interacts with local files but limits its scope to the documentation directory (e.g., docs/roadmap.md, docs/capsule.md). There is no evidence of hardcoded credentials, access to sensitive environment variables, or attempts to read protected system paths.
[REMOTE_CODE_EXECUTION]: No patterns for remote code execution, package installation, or dynamic script generation were detected. The skill instructions focus on text formatting and documentation structure.

roadmap