The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and acts upon data from external API responses and local repository files, establishing a surface for indirect prompt injection. * Ingestion points: External data enters the context via sandbase_run_get (SandBase boundary Step 6) and repository files are inspected during the initial implementation phase (Workflow Step 1). * Boundary markers: There are no explicit instructions for using delimiters or boundary warnings to separate untrusted external content from system instructions. * Capability inventory: The skill provides access to execution tools including sandbase_run, sandbase_discover, and sandbase_inspect to perform automation tasks. * Sanitization: The instructions do not define methods for sanitizing, escaping, or validating the content retrieved from external sources before interpolation into the agent's workflow.

claude-automation-recommender