The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves reading and interpreting content from web applications (DOM, console logs, network headers), which provides a surface for malicious instructions embedded in a target site to influence the agent's behavior.
Ingestion points: Browser-rendered DOM content, console log output, and network request/response data as described in SKILL.md.
Boundary markers: No specific delimiters or instructions to ignore embedded prompts in the analyzed data are defined.
Capability inventory: The skill possesses the ability to interact with UIs (clicks/typing), write files to the local disk (docs/reports/), and navigate to arbitrary URLs.
Sanitization: The skill mandates the redaction of credentials, tokens, and session cookies in references/evidence-bundle.md, mitigating the risk of data exfiltration via the report generation process.
[COMMAND_EXECUTION]: The skill requires starting the application under test and invoking browser automation tools.
Evidence: SKILL.md instructs the agent to "start the application;" using a "start command from the repo/PROJECT_CONTEXT" and to use browser tooling for navigation, interaction, and inspection.

roadtest