skill-chain-prompts
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill implements a workflow management system through instructions and templates. It coordinates existing agent capabilities via a sequential processing model and tracks progress using conversation-native formats like markdown tables and state blocks.
- [INDIRECT_PROMPT_INJECTION]: As an orchestration tool, the skill naturally processes external input (YAML definitions) to determine its workflow, creating an inherent surface for indirect prompt injection.
- Ingestion points: YAML configuration files located in
assets/chains/and user-provided YAML strings through the/skill-chain-prompts runcommand. - Boundary markers: Structured YAML keys (
id,skill,context,args) provide delimitation, although the skill assumes the LLM will correctly parse these as instructions rather than raw data. - Capability inventory: The skill acts as a coordinator and does not possess direct file-system, network, or OS-level capabilities beyond those of the skills it invokes.
- Sanitization: There is no explicit sanitization of the YAML content; the skill relies on the agent's built-in safety mechanisms and instruction-following logic.
Audit Metadata