The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides mechanisms for the agent to store and retrieve external data (vector embeddings and text metadata) to synthesize context for its responses.
Ingestion points: Untrusted data enters the context through database retrieval operations as described in SKILL.md and demonstrated in examples/example-2-multi-database.md.
Boundary markers: The skill documentation explicitly instructs users and the agent to validate and sanitize metadata in SKILL.md ("ALWAYS sanitize metadata to prevent injection attacks").
Capability inventory: The skill has access to Bash, Write, and WebFetch tools as scoped in the frontmatter, and it includes utility scripts for shell-based management.
Sanitization: The skill recommends sanitization but does not provide a standalone sanitization script, relying on the agent's implementation of the provided guardrails.
[COMMAND_EXECUTION]: The skill includes shell scripts for administrative tasks associated with database management.
Evidence: resources/scripts/multi_db_manage.sh uses sqlite3, npx, and standard Linux utilities to initialize shards, perform backups, and optimize databases. These commands are standard for the skill's purpose and align with the permitted Bash tool usage.

agentdb-advanced-features