The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted code from a provided codebase_path to perform a compliance audit.\n- Ingestion points: The skill ingests file content from the codebase_path parameter using grep commands and targeted file reads as defined in the Analysis Workflow and references/factors.md.\n- Boundary markers: The instructions do not define clear delimiters or "ignore embedded instructions" wrappers when processing the codebase files, which is a common vulnerability surface.\n- Capability inventory: The skill uses grep for discovery and deep-dive file reads to gather evidence for generating its report.\n- Sanitization: No sanitization or escaping of the ingested code content is specified before it is processed by the agent, increasing the risk of the agent obeying instructions hidden in the code it audits.

agent-architecture-analysis