code-graph
Warn
Audited by Gen Agent Trust Hub on Jul 5, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill installs the
graphifyypackage from PyPI usinguv,pipx, orpip. - There is a discrepancy between the repository link provided (
github.com/safishamsi/graphify) and the package name specified for installation (graphifyy). - [REMOTE_CODE_EXECUTION]: Provides instructions to install the
uvtool by downloading and executing a shell script directly fromastral.shusingcurl | sh. - [COMMAND_EXECUTION]: Executes several shell commands at runtime to manage the code graph and analyze the repository:
- Runs
graphify .to parse the codebase and generate structural data. - Runs
git log --oneline -30to extract recent commit history for hotspot analysis. - Executes local Python scripts located at
<aiops-root>/skills/aiops/scripts/code_graph_query.pyto perform graph queries. - [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests and processes the entire project codebase (untrusted data) to build the graph.
- Ingestion points: Reads the project directory via
graphifyinbuild-steps.md. - Boundary markers: None identified to separate code content from instructions.
- Capability inventory: Subprocess calls for
graphifyandgit, and execution of local Python scripts. - Sanitization: No explicit sanitization of code content before it is processed by the model for semantic annotation.
Audit Metadata