The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructions and error messages suggest installing the uv tool by piping a script from astral.sh (a well-known service) to the shell: curl -LsSf https://astral.sh/uv/install.sh | sh.
[DYNAMIC_EXECUTION]: The wrapper script scripts/run.py uses uv run to dynamically install and execute the exmergo-dex-core package (a vendor resource) from a remote registry. It also employs os.execvp to perform a process replacement, handing over execution from the Python wrapper to the engine.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources including database schemas, column profiles, and SQL query results.
Ingestion points: Database metadata, column profiles, PII detection results, and raw SQL query results returned by various warehouse connectors (as seen in scripts/run.py and SKILL.md).
Boundary markers: The engine encapsulates all output in a structured JSON envelope. The instructions in SKILL.md explicitly warn against manual schema enumeration.
Capability inventory: The skill can execute SQL queries against connected databases, perform network operations to data warehouses (BigQuery, Snowflake, etc.), and run system commands via its wrapper script.
Sanitization: The skill implements a "query firewall" to block access to PII-flagged columns and uses PII detection algorithms to automatically identify sensitive data during profiling.
[COMMAND_EXECUTION]: The skill invokes the uv package manager and the exmergo-dex-core engine using system calls (subprocess.run, os.execvp) in scripts/run.py.

explore