readability-xai-orchestrator
Pass
Audited by Gen Agent Trust Hub on Apr 6, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. It ingests untrusted text from _workspace/00_input/source.md and passes it to multiple agents (Narrator, Evaluator, Critic) without explicit sanitization or boundary markers defined in the orchestration logic. Malicious instructions within the source text could potentially influence agent behavior. * Ingestion points: Reads source text and reader constraints from the _workspace/ directory (SKILL.md). * Boundary markers: None explicitly defined in the provided orchestration flow. * Capability inventory: File reading/writing in _workspace/, agent creation (TeamCreate), and task management (TaskCreate) in SKILL.md. * Sanitization: No input sanitization or validation logic is present for the ingested source text.
- [COMMAND_EXECUTION]: The skill uses platform-specific multi-agent coordination tools (TeamCreate, TaskCreate, SendMessage) to manage a team of specialized agents. This is the intended functionality for an orchestrator skill.
- [EXTERNAL_DOWNLOADS]: Fetches research references and documentation from arXiv.org, an established repository for scientific preprints.
Audit Metadata