The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted customer messages and ticket descriptions which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: The Usage section in SKILL.md identifies <ticket text, customer message, or issue description> as the primary input source.
Boundary markers: Absent. The instructions do not define specific delimiters or warnings to isolate untrusted input from the agent's instructions.
Capability inventory: The skill has access to sandbase_run, which is used to invoke tools like tavily_search and context_dev_scrape_markdown as mapped in references/sandbase-api-map.md.
Sanitization: Absent. There is no evidence of input validation or escaping before the data is processed by the agent.
[DYNAMIC_EXECUTION]: The skill implements a dynamic discovery and invocation pattern. It uses sandbase_discover to find capabilities, sandbase_inspect to retrieve their schemas, and sandbase_run to execute them based on runtime resolution (SKILL.md, references/sandbase-api-map.md). This mechanism is part of the vendor's platform functionality for resolving support evidence tools.
[EXTERNAL_DOWNLOADS]: The skill fetches content from external sources to support the triage process. It utilizes tavily_search for searching public status pages and documentation, and context_dev_scrape_markdown for reading incident pages and knowledge-base articles (references/sandbase-api-map.md). These operations target well-known services to gather evidence for issue resolution.

ticket-triage