The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data such as system logs and tenant configurations, which presents a surface for indirect prompt injection if those sources contain adversarial instructions.
Ingestion points: Processes logs and configuration files via Read, Glob, and Grep tools.
Boundary markers: The instructions mandate freezing timelines and isolating evidence, though specific text delimiters for untrusted data are not explicitly defined in the prompt structure.
Capability inventory: The skill uses Read, Glob, Grep, WebFetch, Write, and Edit tools to diagnose issues.
Sanitization: The skill contains explicit instructions to "Sanitize evidence while preserving status," "redacted timelines," and "Restrict and redact the bundle before sharing or retaining it" if personal data is present, which significantly mitigates the risk of processing malicious or sensitive content.

workhuman-common-errors