The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves exploring and ingesting content from external web applications, which presents a surface for indirect prompt injection if a target site contains malicious instructions.
Ingestion points: The agent captures page content, metadata, and UI element descriptors using the agent-browser snapshot and agent-browser errors commands.
Boundary markers: The instructions do not specify the use of clear delimiters or instructions to ignore embedded commands within the target application's content.
Capability inventory: The skill possesses the ability to execute shell commands (Bash), perform browser interactions (click, fill, type), and record video/screenshots of the session.
Sanitization: No explicit sanitization or filtering of the ingested web content is described before the data is processed by the agent.
[COMMAND_EXECUTION]: The skill uses local shell commands to manage the testing environment and generate reports.
The agent is instructed to create local directories (mkdir -p) and copy report templates (cp).
These operations are limited to the defined OUTPUT_DIR and utilize local paths relative to the skill directory, representing standard functional behavior for an automated testing tool.

dogfood