The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external sources, including user queries and documentation search results from the Inkeep MCP server, which presents an attack surface for indirect prompt injection.
Ingestion points: The skill retrieves documentation snippets, pricing data, and feature guides through the Inkeep MCP server (mcp__inkeep__*) and processes user-provided integration requirements.
Boundary markers: There are no explicit instructions or delimiters defined to isolate or treat external tool outputs as untrusted, increasing the risk that embedded instructions in the documentation or data could influence the agent's behavior.
Capability inventory: The agent has access to various information tools and can generate complex configuration code, migration scripts, and infrastructure-as-code (Terraform) templates.
Sanitization: No explicit sanitization or filtering logic is present to validate content retrieved from the documentation search before it is incorporated into the agent's reasoning process.
[DYNAMIC_EXECUTION]: The skill provides a technical reference for data migration that includes unsafe deserialization practices.
Evidence: In references/developer-guide.md, the migrate_from_numpy code template uses np.load(metadata_file, allow_pickle=True). Enabling allow_pickle when loading NumPy files can allow for arbitrary code execution if the data file provided to the script is malicious or sourced from an untrusted third party.

ask-zilliz