The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests and processes content from external requirement files. \n
Ingestion points: Requirement documents specified in $ARGUMENTS or the default plan.md file (referenced in SKILL.md). \n
Boundary markers: Absent; the skill does not wrap the file content in delimiters or provide instructions to the agent to ignore embedded commands within those files. \n
Capability inventory: The skill utilizes Bash, Write, EnterPlanMode, and can invoke the planning-with-files skill, providing a path for injected instructions to affect the filesystem or execution environment. \n
Sanitization: Absent; the content read from files is not validated or filtered before being integrated into the agent's context. \n- [COMMAND_EXECUTION]: The skill executes a local Node.js web server (scripts/server.js) via a shell script to provide visual aids during the interview process. \n
The server implementation includes security best practices such as directory traversal protection (using path.basename) and an automatic shutdown mechanism (30-minute idle timeout or parent process exit). \n
While the skill defaults to local binding (127.0.0.1), the start-server.sh script supports binding to all interfaces (0.0.0.0), which is documented as a risk in SKILL.md and should only be used in specific environments.

interview-plan