lark-slides
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided XML content and existing slide data from Feishu/Lark.
- Ingestion points: XML content is read through the scripts/xml_lint.py utility and the lark-cli slides +xml-get command.
- Boundary markers: The skill uses XML tags (e.g., , ) to separate instructions from data content.
- Capability inventory: The skill utilizes lark-cli for slide operations and includes Python scripts that use subprocess.run and os.execv.
- Sanitization: Robust sanitization is present via scripts/xml_lint.py, which validates XML well-formedness and enforces schema constraints using sxsd_validator.py.
- [COMMAND_EXECUTION]: Utility and test scripts within the skill utilize shell command execution for validation and testing.
- Evidence: scripts/iconpark_tool_test.py uses subprocess.run to verify CLI functionality.
- [DYNAMIC_EXECUTION]: The skill contains logic for process replacement and dynamic code compilation in its test and compatibility scripts.
- Evidence: scripts/xml_text_overlap_lint_test.py uses os.execv to chain execution to the main test suite, and scripts/xml_lint_test.py utilizes the compile() function as detected by static analysis.
Audit Metadata