The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill setup requires downloading the @bilig/workpaper package from the NPM registry to provide spreadsheet and formula logic.
[REMOTE_CODE_EXECUTION]: The skill provides configuration examples that use npm exec to download and execute binary tools (bilig-mcp-challenge, bilig-workpaper-mcp) from an external package.
[COMMAND_EXECUTION]: The skill configures the agent to execute shell commands to run an MCP server and perform file system operations on JSON workbook documents.
[PROMPT_INJECTION]: The skill processes spreadsheet formulas and cell data from external files, creating an indirect prompt injection surface if the data source is untrusted.
Ingestion points: Untrusted data enters the agent context via read_cell, read_range, and formula evaluation tools in SKILL.md.
Boundary markers: Absent; while formula validation is suggested, there are no explicit delimiters to prevent the agent from following instructions embedded within cell data.
Capability inventory: The skill possesses the ability to execute shell commands via npm exec and perform file-write operations through export_workpaper_document and serializeWorkPaperDocument in SKILL.md scripts.
Sanitization: The skill provides manual validation advice (rejecting specific shell characters in paths), but lacks automated sanitization for the spreadsheet content or formula results processed at runtime.

bilig-workpaper