The Agent Skills Directory

[SAFE]: The skill defines a legitimate process for automating software development tasks. It coordinates implementation and peer review workflows using standard development practices and referenced internal components.- [PROMPT_INJECTION]: The skill ingests untrusted data from external sources such as issue descriptions and plan files to guide code implementation. While this constitutes an indirect prompt injection surface, no malicious instructions or bypass attempts were identified. Mandatory Evidence: 1. Ingestion points: Issue number/URL, plan file path, and free-text description (SKILL.md); 2. Boundary markers: Absent; 3. Capability inventory: Automated code implementation, branch creation, PR generation, and issue tracking operations (via forge-implement and issue-operations); 4. Sanitization: Not specified.- [DATA_EXFILTRATION]: No patterns for accessing sensitive system files or transmitting data to unauthorized external domains were found. Operations are restricted to standard repository and project management interactions.

forge-ship