The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by ingesting and analyzing untrusted data from repository files.
Ingestion points: The skill is designed to inspect repository files (e.g., READMEs, scripts, configuration files) using Read, Grep, and Glob tools as defined in Step 1 to understand the current workflow state.
Boundary markers: The instructions do not define clear delimiters or specific warnings to ignore instructions embedded within the processed repository files.
Capability inventory: The skill has access to powerful tools including Bash, Write, and Read across the repository scope, allowing it to potentially act on instructions found in files.
Sanitization: There are no explicit steps to sanitize or filter out potential natural language instructions found in the analyzed files before the agent processes them.

workflow-automation