The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill acts as a router for data files, creating a surface where the agent ingests content from external JSON files that could potentially contain instructions. However, the skill lacks any capabilities (network, write, or shell access) that would allow such instructions to be exploited.
Ingestion points: The script scripts/context_router.py discovers and reads JSON files located in the .agent/skills/*/data/ directory structure.
Boundary markers: The script outputs the filtered JSON data directly to the agent's context without surrounding delimiters or 'ignore' instructions.
Capability inventory: The script is limited to standard library file operations and keyword filtering; it does not invoke subprocesses, perform network requests, or modify the file system.
Sanitization: No sanitization is performed on the data retrieved from the JSON files.

context-router