The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill processes file paths and linting error messages from the local filesystem and interpolates them directly into prompts for sub-agents via the Task tool.\n
Ingestion points: analyze_errors.sh extracts file paths from pnpm lint output, which are then passed to sub-agents via the Task tool prompt.\n
Boundary markers: The prompts generated for sub-agents in Step 4.2 lack explicit delimiters or instructions to ignore embedded instructions within the file list or error descriptions.\n
Capability inventory: The skill and its sub-agents have significant capabilities, including code modification (Edit), shell execution (Bash), and agent orchestration (Task).\n
Sanitization: No sanitization or escaping is performed on the extracted filenames or error messages before they are interpolated into the natural language prompts for the Task tool.\n- [Command Execution] (LOW): The analyze_errors.sh script uses xargs without the -0 flag (null delimiter), making it sensitive to filenames containing spaces or shell metacharacters.\n
Evidence: The command echo "$FILES_WITH_ERRORS" | xargs -n1 dirname will incorrectly split filenames containing spaces, leading to potential script failure or the processing of unintended directory paths if a project contains maliciously named files.

fix-eslint