Security audit
Security checks for vulnerabilities and agentic risk
This skill has documentation and implementation mismatches, but its behavior is limited to local, user-directed data conversion with no evidence of hidden exfiltration, persistence, or destructive activity.
Before installing, treat this as a basic local format-conversion utility rather than the broader text transformation tool described in SKILL.md. Review input and output paths carefully, because it can read a chosen file and overwrite or create the chosen output file.
The documented purpose and the detected behavior do not align: the skill claims broad text transformation features, while analysis indicates undeclared file access and different conversion behavior than advertised. This mismatch is security-relevant because users and orchestrators may grant trust or invoke the skill under false assumptions, leading to unsafe handling of local files or incorrect processing outcomes.
The skill advertises file-oriented usage (transform-tool [options] <file>) and example commands that read from local files, but it declares no permissions or allowed-tools scope. That creates an authorization/documentation gap: consumers and platforms cannot accurately constrain or review the skill's file read/write behavior, increasing the chance of unintended local file access or modification.
The manifest describes a broader transformation tool for case changes, encoding conversion, and data cleaning, but the code implements only CSV/JSON/YAML/XML/TSV conversion paths. There is no functionality for case transformation, character encoding conversion, or general text-cleaning operations, so the stated behavior does not match the actual implemented scope.
The argument help says the target format can be 'csv, json, xml, yaml, tsv', but convert() only supports json, csv, tsv, and yaml as targets and raises an error for xml. This is an active contradiction between the user-facing documentation/help and the actual code behavior.
No suspicious patterns detected.
