writing-shape
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from a 'raw material markdown file' and incorporates this content into an article file. This processing of external data, combined with file read and write capabilities, creates a surface for indirect prompt injection where instructions embedded in the raw material could influence agent behavior.
- Ingestion points: The skill is instructed to 'Read the pile' from an external markdown file at the beginning of the session.
- Boundary markers: The instructions do not specify the use of delimiters or 'ignore' directives to separate the raw material data from the agent's system instructions.
- Capability inventory: The skill has permissions to read files ('Read the pile') and perform sequential file writes ('Append to the article file as you go') to the local file system.
- Sanitization: There is no requirement in the skill instructions to sanitize, escape, or validate the content extracted from the raw material before it is used to generate article text.
Audit Metadata