The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute goldsky secret commands and a local helper script scripts/parse-connection-string.sh to manage credentials and parse connection strings. These are standard operations for the skill's stated purpose.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data in the form of connection strings provided by the user or external sources. While it includes instructions to show the user the parsed data and mask passwords, there is an inherent risk that a malicious connection string could contain instructions intended to influence the agent's behavior.
Ingestion points: Connection strings are ingested in Step 3 of SKILL.md and processed by scripts/parse-connection-string.sh.
Boundary markers: The skill lacks explicit boundary markers or instructions to ignore embedded commands within the connection strings, although it does recommend user confirmation.
Capability inventory: The agent can execute shell commands (goldsky secret list, create, reveal, update, delete) and run a local bash script.
Sanitization: The parse-connection-string.sh script performs basic parsing and JSON escaping but does not fully sanitize against NL-based prompt injection.

secrets