The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on various system utilities to perform its operations. This includes psql for database queries, pg_dump and pg_restore for backup management, git for repository checks, curl for fetching online documentation, and python3 for internal logic and TOML profile management.\n- [DATA_EXFILTRATION]: The scripts/bootstrap_profile.py utility scans the project directory for sensitive configuration files such as .env, .json, .yaml, and .toml to identify existing database connection strings. While this is an intended feature to simplify setup, it involves automated access to sensitive local file paths.\n- [EXTERNAL_DOWNLOADS]: The scripts/search_postgres_docs.sh script performs network requests to the official PostgreSQL documentation website (postgresql.org) to provide documentation search results. This is a download from a well-known and trusted official service.\n- [PROMPT_INJECTION]: The skill processes untrusted data from external sources, specifically database query results and documentation search results, which could contain malicious instructions designed to influence the agent's behavior (Indirect Prompt Injection surface).\n
Ingestion points: Results from SQL queries executed via scripts/run_sql.sh and documentation snippets fetched from postgresql.org via scripts/search_postgres_docs.sh.\n
Boundary markers: Absent. No explicit delimiters or instructions to ignore embedded commands are used when presenting external data to the agent context.\n
Capability inventory: The skill has broad capabilities including arbitrary SQL execution, file system modification (TOML and CHANGELOG updates), and network access via curl.\n
Sanitization: No specific sanitization or filtering of database output or documentation snippets for malicious prompt content is performed.

postgres