The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The scripts/publish.sh script contains a recommendation in an error message to execute curl -fsSL https://here.now/install.sh | bash if the jq dependency is missing from the host system. This pattern involves downloading and executing unverified remote code directly in the shell. While the command is not executed automatically, an AI agent encountering the error might attempt to run the suggested command to fix the environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes arbitrary files and directories provided by the user for publication. This creates an attack surface where malicious instructions embedded in those files could influence the agent's behavior when it reads or prepares them for upload.
Ingestion points: scripts/publish.sh reads the contents of the file or directory specified in the TARGET argument.
Boundary markers: No specific delimiters or safety warnings are applied to the content of the files being processed.
Capability inventory: The skill possesses capabilities for file system access (read/write), directory traversal, and network operations via curl to the here.now API.
Sanitization: The script performs hashing and uploads file content without any sanitization or validation of the data itself.
[COMMAND_EXECUTION]: The skill relies on a bash script (scripts/publish.sh) that executes multiple shell commands and network requests. This reliance on the local shell environment and external binaries like curl, file, and jq introduces a dependency on the integrity of those tools and the overall system environment.

here-now