skill-cross-platform-publish
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from a JSON manifest for adaptation to multiple platforms. Ingestion points: The content field (title, body, tags) in the manifest JSON, processed by the LLM as instructed in SKILL.md. Boundary markers: No explicit delimiters are used to separate user content from agent instructions, which could allow embedded instructions to influence the LLM. Capability inventory: The skill routes data to multiple publisher skills (e.g., skill-xhs-publisher) and executes logging scripts. Sanitization: No sanitization or filtering of input content is performed before processing.
- [COMMAND_EXECUTION]: The skill executes scripts/publish_dispatch.py to validate constraints and generate plans, which involves reading manifest files from the local filesystem using standard Python libraries.
Audit Metadata