The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external data from call transcripts and emails, which constitutes a potential surface for indirect prompt injection attacks.
Ingestion points: Processes content from calls, emails, and deals as described in references/task-method.md.
Boundary markers: The skill instructions emphasize structural adherence to templates (objections-output.md, overview-output.md) but do not explicitly instruct the agent to ignore hidden commands within the ingested transcripts.
Capability inventory: The skill has permissions to execute workspace mutations and public hosting operations as outlined in SKILL.md.
Sanitization: No specific data sanitization or escaping mechanisms for external transcript data are documented.
[COMMAND_EXECUTION]: The skill utilizes platform-provided tools to interact with the environment and process data.
Evidence: References to workspace-mutations.md and host-runtime.md indicate the skill is configured to use authorized platform tools for data retrieval and state management.
[EXTERNAL_DOWNLOADS]: The skill references shared configuration and logic files using relative paths within the platform's file system.
Evidence: Accesses files such as ../shared/gtm-context.md, ../shared/output-readiness.md, and ../../shared/analytics-contract.md. These are vendor-owned shared resources rather than untrusted external downloads.

insights