Security audit
Security checks for vulnerabilities and agentic risk
This skill is not malicious, but it asks for broad file, execution, network, and output authority with triggers and retention details that are too broad for automatic use.
Review before installing. Use this only if you are comfortable with a broad document processor that can read many file types, run parsing code, create outputs, recurse through archives, and potentially use web or external transcription services. Avoid using it on confidential files unless network use, output locations, and cleanup/retention behavior are clear and controlled.
The auto-activation triggers are extremely broad phrases like 'read this', 'parse this', 'convert to', and 'summarize this', which overlap with common benign user requests. This can cause the skill to activate unexpectedly, invoking powerful file, archive, OCR, transcription, code execution, or network-enabled behavior in contexts where the user did not intend to use this high-capability skill.
The skill workflow explicitly contemplates adding context from external sources, and the manifest includes network tools, but the user-facing description does not clearly warn that processing may involve external network access. Users may therefore provide sensitive documents without understanding that content could be transmitted or influenced by remote resources.
The generate_image capability is unrelated to the stated role of universal file parsing and synthesis, violating least privilege. While less severe than network access, unnecessary output-generation capabilities can be abused to create deceptive artifacts or expand the operational scope beyond what users expect from a file-processing skill.
The generate_image capability is unrelated to the stated role of universal file parsing and synthesis, violating least privilege. While less severe than network access, unnecessary output-generation capabilities can be abused to create deceptive artifacts or expand the operational scope beyond what users expect from a file-processing skill.
The skill says source files are read-only and must never be modified, but elsewhere declares write_file and edit_file capabilities. This mismatch creates ambiguity for the agent and reviewers, increasing the risk that original files or extracted contents could be altered despite the stated safety rule.
The output delivery section documents generated files and archive extraction behavior, but the skill does not clearly warn users up front that it may create output artifacts, temporary extracted data, or intermediate files. For sensitive inputs, this can lead to unintended persistence of confidential material in the workspace or output paths.
No suspicious patterns detected.
