The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes a shell script located at scripts/snapshot-push.sh to perform repository operations. While the script is local to the project, this pattern represents the execution of arbitrary system commands.
[INDIRECT_PROMPT_INJECTION]: The skill accepts an optional message input from the user which is directly interpolated into a shell command: bash scripts/snapshot-push.sh "<optional message>". This creates a vulnerability surface where a malicious user could attempt to inject additional shell commands if the agent does not properly escape the input.
Ingestion points: The message field provided by the user in the input contract.
Boundary markers: Absent. The instruction does not specify delimiters or escaping requirements for the interpolated message.
Capability inventory: Execution of shell scripts via bash in the SKILL.md workflow.
Sanitization: Absent. There are no instructions to sanitize, validate, or escape the user-provided message before it is passed to the shell.

snapshot-push