The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes raw, untrusted data from customer support tickets, which creates an inherent surface for indirect prompt injection. A malicious user could attempt to include instructions within a ticket to influence how the agent classifies or routes the request.
Ingestion points: The workflow in SKILL.md describes ingesting raw subject lines, bodies, and customer metadata.
Boundary markers: There are no explicit instructions or delimiters defined to isolate the ticket content from the agent's logic.
Capability inventory: The agent generates responses and determines routing paths for the tickets.
Sanitization: The provided instructions do not mention sanitizing or validating the input text to prevent instructional hijacking.
[NO_CODE]: This skill is composed entirely of markdown instructions and does not bundle any scripts, binaries, or configuration files that could execute code.
[SAFE]: No evidence of malicious behavior, such as credential exposure, unauthorized network connections, or persistence mechanisms, was found in the analysis of the skill documentation.

ticket-triage