The Agent Skills Directory

[SAFE]: The skill follows a well-defined requirements discovery process. It scans project-specific files (e.g., AGENTS.md, CLAUDE.md) to gather context but does not access sensitive system directories or credentials.
[SAFE]: The execution flow includes an explicit manual confirmation step (Phase 6) where the user must approve the generated specification card before it is considered 'Confirmed'. This serves as a significant security gate against unintended actions.
[SAFE]: Tool usage (e.g., question for user interaction and webfetch for information gathering) is consistent with the skill's stated purpose of requirement elicitation and clarification.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied information and project files to generate instructions for downstream skills. While this creates an attack surface for indirect prompt injection (e.g., malicious instructions hidden in a project's handoff file), the requirement for explicit user review and the use of a structured specification template with 'Excluded Items' effectively mitigates this risk.
[SAFE]: No obfuscation, remote code execution, or persistence mechanisms were found. The installation instructions provided in the README are for the user to perform manually via standard git commands.

rdq-opencode