The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill reads and processes external source files to update documentation, creating a surface for injection.
Ingestion points: Contents of source files specified in .docstore/sources.yaml.
Boundary markers: None specified in the workflow instructions.
Capability inventory: Bash, Read, Write, Edit, Grep, Glob, Task.
Sanitization: No explicit sanitization or filtering of source content is mentioned before extraction or integration.
[Command Execution] (SAFE): Uses the Bash tool to execute stat for retrieving file modification timestamps.
Evidence: stat -f %m <file> (macOS) and stat -c %Y <file> (Linux) are used for legitimate metadata comparison.
[Data Exposure & Exfiltration] (SAFE): Accesses local file paths for document management. No hardcoded credentials or network-based exfiltration patterns were detected.

doc-update