transition-weaver
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes project configuration files to generate prose, creating a surface for indirect prompt injection.
- Ingestion points:
outline/outline.ymlandoutline/subsection_briefs.jsonlfiles are read from the workspace. - Boundary markers: Data files do not use explicit delimiters to isolate untrusted content from the processing logic.
- Capability inventory: The skill script is restricted to writing the
outline/transitions.mdfile within the local workspace; it lacks network access and command execution capabilities. - Sanitization: Input data is parsed using
yaml.safe_load()and cleaned of excessive whitespace. - [SAFE]: No malicious patterns such as prompt injection, obfuscation, or data exfiltration were detected. The skill uses standard local file operations and provides security-conscious advice, such as recommending ASCII characters over Unicode to prevent character-encoding attacks.
Audit Metadata