The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and analyze untrusted frontend source code and build artifacts. This creates a surface for indirect prompt injection, where malicious instructions hidden in code comments or project metadata could attempt to override the agent's behavior during a review.\n- Ingestion points: The skill directs the agent to search through files in the src/, app/, public/, and build output directories (dist/, build/, out/).\n- Boundary markers: No specific instructions are provided to the agent to use delimiters or to disregard instructions found within the data being analyzed.\n- Capability inventory: The skill suggests using shell-based search tools like rg and find to probe the codebase, which can be influenced by untrusted filenames or file contents.\n- Sanitization: There are no requirements defined for the agent to sanitize or validate the external project data before it is processed.

client-error-observability-contracts