The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted tabular data from external CSV and Excel files, which is subsequently summarized and interpreted by the AI agent. This creates a surface for indirect instructions embedded in the data to influence agent behavior. • Ingestion points: src/data_analyzer.py loads and processes user-provided data files. • Boundary markers: No explicit delimiters or warnings to ignore instructions within the data are present in the skill manifest. • Capability inventory: The skill has the ability to write to the file system, execute shell commands, and control browser instances. • Sanitization: While data is parsed into DataFrames, the resulting values are interpolated into summaries for AI interpretation without specific NLP-level sanitization.
[COMMAND_EXECUTION]: The skill performs system-level command execution to handle data processing and automated testing. • src/index.js utilizes execFile to execute Python code snippets for tasks such as extracting Excel sheet names. • tests/test_pipeline.py employs subprocess.run to execute CLI commands for pipeline verification.
[EXTERNAL_DOWNLOADS]: The skill downloads browser binaries as part of its setup and operational flow. • Fetches Chromium binaries from Microsoft's official servers to support the Playwright-based screenshot generation functionality used in creating reports.

data-analyze-refined