compiler

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill explicitly requires copying verbatim source lines and transcribing code/files into the artifact (e.g., "copy the matched line verbatim", "# Grounding: transcribed", and re-open cited file:line to confirm quotes), which forces any secrets present in inputs to be included unchanged in the LLM's outputs.

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

• Third-party content exposure detected (high risk: 0.75). The required workflow reads “ANY research input” (including PDFs/repos/logs/notes/threads) and then “READ all inputs thoroughly” and “re-read the source” to extract and transcribe tables/figures/text into the agent’s LLM context, so if the provided inputs include outsider-authored free text (e.g., a public repo issue/wiki/discussion or downloaded document), that text can be ingested at runtime.