The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute high-privilege and potentially destructive shell commands. Specifically, references/replication.md includes commands to stop the database service (systemctl stop postgresql) and delete the primary data directory (rm -rf /var/lib/postgresql/14/main/*) to facilitate replication setup. It also advocates for the use of the pg_repack CLI for table maintenance.- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and processes data from PostgreSQL system views that aggregate input from external database clients.
Ingestion points: Data is ingested from pg_stat_statements, pg_stat_activity, and pg_stat_user_tables as seen in SKILL.md and references/maintenance.md.
Boundary markers: Absent; the instructions do not implement markers to distinguish between data and potential commands within the query logs or activity summaries.
Capability inventory: The skill possesses the capability to execute shell commands and administrative SQL functions like pg_terminate_backend.
Sanitization: No evidence of filtering or sanitization of database metadata is found.- [METADATA_POISONING]: The skill's YAML frontmatter in SKILL.md lists the author as 'Jeffallan', which conflicts with the identified author 'seltherpython'. This inconsistency in metadata could be used to misrepresent the provenance of the skill.

postgres-pro