The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple bash commands (roborev, git, go test) and incorporates user-provided job IDs and tool-generated metadata into these command strings. This creates a potential surface for command injection if inputs containing shell metacharacters are processed without sufficient sanitization.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection by ingesting and acting upon review findings from an external tool. \n- Ingestion points: Findings are parsed from the output field of the roborev show command as described in SKILL.md. \n- Boundary markers: The skill definition does not specify the use of boundary markers or explicit instructions for the agent to ignore instructions embedded within the ingested findings. \n- Capability inventory: The agent is granted capabilities to modify the local file system (to apply fixes) and execute a variety of shell commands. \n- Sanitization: While the skill includes an instruction to escape quotes for the roborev comment command, it lacks a comprehensive sanitization strategy for all data ingested from the review output.

roborev-fix