The Agent Skills Directory

[SAFE]: No malicious patterns or security vulnerabilities were detected. The skill logic is focused on interactive requirements gathering and includes explicit restrictions against sensitive operations like git commands or plan generation.
[COMMAND_EXECUTION]: The skill orchestrates specialized sub-agents (Explore, interviewer) using the Agent tool to perform structured questioning and context gathering based on user input.
[DATA_EXFILTRATION]: Accesses local codebase patterns using an 'Explore' sub-agent when the --deep flag is enabled. This data informs the interview and is stored in a local insights file; no network exfiltration is attempted.
[PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection through its ingestion of untrusted data.
Ingestion points: User-provided interview answers and codebase context extracted via sub-agents (SKILL.md).
Boundary markers: None explicitly defined in sub-agent prompts to distinguish between instructions and data.
Capability inventory: File writing (Write tool) and sub-agent orchestration (Agent tool).
Sanitization: No explicit sanitization or filtering is specified for user inputs or codebase findings before they are passed to the interviewer agent.

deep-interview