writing-shape

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill requires the agent to read a user-supplied "pile" and to "pull a fragment" or "quote when the original wording is the point," which can force the model to reproduce input text verbatim — including any embedded secrets or credentials.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). The required workflow explicitly instructs the agent to “Read the input file in full” and then “mine the pile to fill” the article, meaning the LLM ingests the user-supplied markdown raw material (outsider-authored free text if the operating user didn’t author it).