The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to execute shell commands within the remote Antithesis test environment and JavaScript within the browser context to automate UI interactions.
Evidence: The skill uses agent-browser eval to execute functions from the provided assets/antithesis-debug.js runtime. It also uses the debugger's native capability to run bash scripts in target containers, as described in references/simplified-debugger.md and references/common-inspections.md.
[EXTERNAL_DOWNLOADS]: The skill provides a mechanism for downloading files from the remote Antithesis debugging environment to the local filesystem.
Evidence: references/simplified-debugger.md details a workflow using agent-browser download to retrieve artifacts extracted by the debugger.
[DATA_EXFILTRATION]: The skill includes functionality to "Extract file" and download data from the containerized Antithesis environment to the user's local machine for analysis.
Evidence: Found in the extraction workflows described in SKILL.md and references/simplified-debugger.md.
[PROMPT_INJECTION]: The skill ingests untrusted data from the Antithesis debugger UI, such as log rows and container metadata, creating a surface for indirect prompt injection.
Ingestion points: Log view browsing and event stream analysis in references/simplified-debugger.md and references/common-inspections.md.
Boundary markers: No explicit boundary markers or delimiters for ingested content are specified in the instructions.
Capability inventory: The skill can execute arbitrary shell commands in remote containers and run JavaScript in the browser context via agent-browser eval.
Sanitization: No explicit sanitization or validation of the ingested external content was observed before it is processed by the agent.

antithesis-debug