cognee-switch-datasets
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes multiple local Python and shell scripts to manage Cognee session state and datasets. These include
switch-dataset.py,cognee-doctor.sh,sync-session-to-graph.py, andcognee-search.sh, all located within the${CLAUDE_PLUGIN_ROOT}directory. - [INDIRECT_PROMPT_INJECTION]: The skill instructions allow the interpolation of untrusted user input directly into shell command templates, creating a potential command injection surface.
- Ingestion points: User-controlled content enters the skill via the
$ARGUMENTSvariable and through the "Other" text input option in theAskUserQuestiontool as described inSKILL.md. - Boundary markers: The skill uses double quotes (e.g.,
"$ARGUMENTS"and"<chosen name>") when calling scripts. While this provides basic protection against simple argument splitting, it does not prevent all forms of shell metacharacter exploitation. - Capability inventory: The skill possesses the capability to execute shell commands and modify local session management records via the provided plugin scripts.
- Sanitization: The instructions do not define any validation, sanitization, or character-set restrictions for the dataset names provided by the user before they are passed to the command-line interface.
Audit Metadata