The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions require the agent to ingest and analyze data from external database statistics views, which could theoretically be manipulated to include malicious instructions.
Ingestion points: SQL query results from pg_stat_user_tables, pg_stat_statements, and other database metadata views as described in SKILL.md.
Boundary markers: Absent; there are no specific instructions or delimiters to isolate the data from the agent's instruction stream.
Capability inventory: The skill has access to database CLIs (psql, mysql, mongosh) and file system operations (Read, Write, Edit).
Sanitization: Absent; the agent is instructed to directly interpret and act upon the output of database queries.
[METADATA_POISONING]: There is a significant discrepancy between the description of the scripts/analyze_indexes.py script and its actual code implementation. The scripts/README.md and the script's own docstring claim it performs database index analysis, but the code is actually a generic directory and file size analyzer.
[COMMAND_EXECUTION]: The skill uses Bash tools scoped to database clients (psql, mysql, mongosh) to perform its primary function of index management. While powerful, this behavior is consistent with the skill's stated purpose of database optimization.

analyzing-database-indexes