The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by processing untrusted data from external Clay tables.
Ingestion points: Data enters the context via the table MCP tool and the clay tables query and clay tables columns get CLI commands as described in SKILL.md and dependency-catalog.md.
Boundary markers: There are no explicit instructions or delimiters defined to isolate the retrieved table data or warn the agent to ignore instructions embedded within values or formulas.
Capability inventory: The skill has the ability to execute shell commands (via the clay CLI) and write data to the local filesystem (CSV export instructions in SKILL.md).
Sanitization: The skill suggests basic CSV quoting for formatting but lacks semantic validation or filtering of the content retrieved from the external source.
[COMMAND_EXECUTION]: The skill relies on the execution of the clay CLI and other shell utilities like jq to perform its core functions. It provides multiple bash examples for listing tables, updating query states, and running complex queries through pipes. This behavior is documented in SKILL.md and dependency-catalog.md.

tables