The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous SQL templates and workflows that instruct the agent to query database system catalogs such as pg_stat_activity, pg_stat_statements, pg_locks, and pg_stat_io. These operations are standard for performance tuning and database diagnostics and are used safely within the skill's defined scope.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, untrusted data including SQL schema files, query plans, and DDL scripts. It includes an inherent attack surface for indirect prompt injection; however, this is addressed by clear 'Gates' in each workflow which require the agent to verify performance claims (e.g., comparing buffer counts in EXPLAIN ANALYZE) and constraint integrity (pg_index.indisvalid) rather than relying on user-provided assertions.
[EXTERNAL_DOWNLOADS]: The skill's documentation (NOTICE.md and SOURCES.yaml) cites official and reputable sources including the PostgreSQL documentation and public repositories from well-known organizations such as Microsoft, Supabase, and Neon. These references are used for attribution and technical content merging, and no runtime execution of untrusted remote code was identified.

postgres