The Agent Skills Directory

[COMMAND_EXECUTION]: The validation step for JSON5 files is vulnerable to both shell command injection and JavaScript injection. The variable $SPEC, which contains the file path, is interpolated into a node -e shell command. An attacker can use shell metacharacters (e.g., backticks) or JavaScript string delimiters (single quotes) within a filename to execute arbitrary code during the validation process.
[EXTERNAL_DOWNLOADS]: The skill relies on the editly CLI tool being pre-installed on the system and assumes the availability of the json5 Node.js package. These external dependencies are not managed or verified by the skill logic.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting untrusted data from the JSON specification file ($SPEC). There are no boundary markers or delimiters used to separate the specification content from the agent's instructions. The skill possesses significant capabilities, including arbitrary shell command execution (Bash) and file system writes (Write). No sanitization, validation, or filtering of the specification content is performed before processing, which could allow instructions embedded in the file to influence the agent's behavior.

editly-render