The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs directory scanning and file system modifications, specifically using atomic write patterns involving the mv command to ensure the integrity of the tracking and configuration files.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it parses the front-matter of untrusted markdown files within application directories to determine the status of the job pipeline.
Ingestion points: Reads data from tracker.yaml, .jobops/config.json, and scans the first 30 lines of arbitrary .md files found in subdirectories of the application root.
Boundary markers: Does not employ delimiters or specific instructions to ignore embedded commands within the files being scanned.
Capability inventory: Possesses filesystem read/write permissions and the ability to programmatically invoke other skills in the JobOps suite.
Sanitization: Metadata extracted from markdown files is used to set internal flags without explicit validation or sanitization of the content.

dashboard