The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from external sources and interpolates it into an HTML document. \n
Ingestion points: Data is gathered from CRM records and call transcripts using tools like search_call_transcripts and list_findings (referenced in references/tool-reference.md). \n
Boundary markers: Absent. There are no instructions to use delimiters or ignore instructions within the ingested data. \n
Capability inventory: The skill uses a wide range of read-only tools to access deal events, findings, and knowledge base entities. \n
Sanitization: Absent. The instructions in references/html-architecture.md and references/sections.md do not include requirements for escaping or sanitizing content retrieved from transcripts before including it in the report. \n- [EXTERNAL_DOWNLOADS]: The skill uses external font providers for styling the report. \n
Evidence: references/html-architecture.md allows for links to well-known services like Google Fonts and Typekit. \n- [COMMAND_EXECUTION]: The skill invokes various platform-integrated tools to perform its analytical tasks. \n
Evidence: Tools such as list_events, get_event_detail, and search_call_transcripts are used to query external data sources.

win-loss-report