The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted local data (existing tsconfig.json and package.json files) during its initialization phase.
Ingestion points: The explore-codebase sub-agent reads local configuration files to determine the project runtime and existing settings.
Boundary markers: No specific delimiters or "ignore instructions" warnings are defined for the read operation.
Capability inventory: The skill is capable of modifying filesystem configuration and querying external documentation via sub-agents.
Sanitization: The output is restricted to structured JSON configuration following the TypeScript schema, which inherently limits injection vectors.
[EXTERNAL_DOWNLOADS]: The skill references official and well-known documentation sources.
Fetches current flag behaviors from the official Microsoft TypeScript GitHub repository via the query-docs tool.
References runtime documentation from established domains including typescriptlang.org, bun.sh, and nodejs.org.
[SAFE]: The skill demonstrates standard development utility behavior. All sub-agents (fuse-ai-pilot) and referenced tools are consistent with the author's (fusengine) ecosystem. No patterns related to credential harvesting, persistence, or command injection were found.

ts-config