The Agent Skills Directory

Indirect Prompt Injection Surface: The skill processes untrusted external data from sources like product requirement documents (PRDs) and tracker comments. 1. Ingestion points: External documentation, Drive files, and ticket history are loaded into the context for review (SKILL.md). 2. Boundary markers: The instructions do not define explicit delimiters or instructions to ignore embedded prompts within the ingested documents. 3. Capability inventory: The skill interacts with local configuration files, writes to specific matter folders, and posts to tracker APIs (SKILL.md). 4. Sanitization: There is no evidence of automated sanitization or filtering of the ingested external content. While this is inherent to a document review workflow, it creates a surface where malicious instructions in a document could attempt to influence the agent's analysis.

launch-review