The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external data files (Excel, CSV) provided by users and includes the capability to write reports to the local filesystem. This creates a potential surface where malicious instructions embedded in project data could influence agent behavior during processing.
Ingestion points: The skill loads project data using pd.read_excel as shown in SKILL.md and supports CSV/JSON formats per instructions.md.
Boundary markers: The instructions.md file includes a directive to "Validate inputs before processing," which acts as a mitigation strategy.
Capability inventory: The skill possesses file-writing capabilities through plt.savefig and PdfPages in SKILL.md, and references filesystem permissions in claw.json.
Sanitization: The agent is instructed to validate inputs and report errors, providing a baseline level of protection against malicious or malformed data.

data-visualization