The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill utilizes automation to generate TypeScript modules and HTML/JSON documentation from source data files, which presents a risk of arbitrary code or data modification if source TSVs are compromised.
The script scripts/crosswalk/gen.py is documented as a tool that generates TypeScript modules from research TSV files.
The script scripts/atlas/build.ts dynamically generates a static HTML atlas that embeds complex JSON structures using JSON.stringify on local data objects.
[INDIRECT_PROMPT_INJECTION]: The skill architecture relies on fetching and processing external state education standards which creates a significant surface for indirect injection attacks.
Ingestion points: Data enters the agent's context through research TSVs located in the data/ directory (e.g., data/crosswalks/*.tsv).
Boundary markers: The skill lacks explicit boundary markers or instructions to treat ingested standards data as untrusted, increasing the likelihood that the agent might obey malicious instructions embedded in a state's standard summary.
Capability inventory: The skill possesses the capability to perform file system operations (read/write) and execute CLI tools via scripts/atlas/build.ts.
Sanitization: While scripts/atlas/build.ts includes an esc() function for basic HTML entity replacement, there is no verification or sanitization of the natural language summaries processed from the standards database.
[COMMAND_EXECUTION]: The skill contains a build script that executes commands on the host environment.
The Bun script scripts/atlas/build.ts uses Bun.spawnSync to execute npx oxfmt. While this specific command is a standard utility, the use of process.argv to determine the target path could be misused to target or format arbitrary files on the system if not used carefully by the developer.

us-math-standards