The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses sudo with system package managers (e.g., apt-get, dnf, pacman) to install the grpcurl utility during prerequisite setup.
[COMMAND_EXECUTION]: The skill captures output from a grpcurl command into a shell variable and interpolates that variable directly into the JSON payload of a subsequent shell command. This pattern creates a shell injection risk if the data returned by the server contains malicious characters intended to break out of the string context.
[EXTERNAL_DOWNLOADS]: Provides instructions to download grpcurl binaries from the official FullStory GitHub repository.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted data from external gRPC endpoints.
Ingestion points: Data retrieved from gRPC server methods (stored in variables like RESULT and VALUE).
Boundary markers: No boundary markers or delimiters are used to differentiate untrusted server data from the agent's instructional context.
Capability inventory: The skill possesses the capability to execute shell commands (grpcurl, curl, jq) and write files to /tmp.
Sanitization: There is no evidence of sanitization, escaping, or validation of the variables extracted from gRPC responses before they are used in downstream shell execution.

seed-hypermedia-read