The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides instructions for handling and testing untrusted data retrieved from external sources, which constitutes a vulnerability surface for indirect prompt injection.
Ingestion points: The skill processes data from AI_RUN_MANIFEST.yaml, HUMAN_OVERSIGHT_CASE_<task_id>.yaml, and untrusted content from tools, MCP, and A2A (agent-to-agent) handoffs as defined in docs/agile-v-runtime/05_AGENT_TOOL_AND_DELEGATION_CONTRACT.md.
Boundary markers: The instructions explicitly command the agent to treat retrieved content as "untrusted data" and ensure it is "quoted/treated as data" to prevent execution of embedded instructions.
Capability inventory: The skill is primarily focused on generating text-based test specifications (TC-XXXX); it does not include commands for sensitive file modification or unauthorized network access.
Sanitization: The skill mandates specific "Agentic Security Tests" designed to prove that external content cannot grant unauthorized authority or trigger undeclared side effects.
[NO_CODE]: The skill consists exclusively of instructional Markdown and YAML metadata without any accompanying scripts or executables.

test-designer