The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it parses user-provided spreadsheets and displays cell samples to the agent. Maliciously crafted content in a spreadsheet could potentially attempt to influence the agent's subsequent actions.\n
Ingestion points: scripts/analyze.py reads data from user-supplied CSV and XLSX files.\n
Boundary markers: The agent receives the data in a structured Markdown report, but there are no explicit delimiters instructed to the agent to treat this specific content as untrusted data.\n
Capability inventory: The agent has the ability to run local scripts (analyze.py, apply.py) which can read and write files.\n
Sanitization: Data is processed through the pandas library, but the final interpretation depends on the agent's prompt handling logic.\n- [EXTERNAL_DOWNLOADS]: The skill's setup process involves downloading standard, well-known data processing libraries (pandas, openpyxl) from official package registries.\n- [COMMAND_EXECUTION]: The agent executes local Python scripts provided within the skill's own repository to perform analysis and file transformations. These scripts operate locally on the user's filesystem and do not perform network operations.

biohousekeeper