The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill interacts with external product interfaces that may contain adversarial instructions, creating a surface for indirect prompt injection. 1. Ingestion points: Data enters the agent's context through browser snapshots and text extraction from product surfaces via the agent-browser tool, as well as local documentation in the qa-docs-path. 2. Boundary markers: The persona-fidelity documentation provides behavioral guidance to avoid evaluator framing, but no formal delimiters or instructions to ignore embedded commands in ingested data are present. 3. Capability inventory: The skill can write report files, capture screenshots, and modify product source code through its fix-loop mechanism. 4. Sanitization: The skill does not describe any specific validation or sanitization routines for content ingested from the browser or local documentation.
[COMMAND_EXECUTION]: The skill relies on the agent-browser tool for UI interaction and references shell-based activities for regression testing and applying bug fixes to the codebase.
[DYNAMIC_EXECUTION]: The Step 7 Fix Loop authorizes the agent to autonomously modify source code files to resolve identified bugs. While this is restricted by specific bounds (e.g., small, well-understood, low-risk fixes), it involves dynamic code generation and modification of the local product environment.

qa-execution