writing-shape
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process raw, unstructured markdown data from an external file (referred to as "the pile"). This creates a surface for indirect prompt injection, where instructions embedded within the raw material could potentially influence the agent's behavior or be executed as commands if not properly delimited.
- Ingestion points: The instructions in
SKILL.mddirect the agent to "Read the input file in full" and "mine the pile" to generate content. - Boundary markers: There are no explicit delimiters or instructions to the model to ignore potential commands or formatting instructions embedded within the raw input material.
- Capability inventory: The agent is instructed to "Append to the article file as you go," which involves writing content derived from the potentially untrusted input pile to a local file.
- Sanitization: The skill lacks any explicit sanitization, filtering, or validation steps for the content extracted from the raw material before it is used to generate the final article.
Audit Metadata