The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script located at <plugin root>/scripts/prose-lint.py as part of its reporting workflow in SKILL.md.
[COMMAND_EXECUTION]: When testing CLI tools, the skill is instructed to invoke the built binary in a shell as described in references/surfaces.md. This involves executing external code that may be untrusted if the product being tested is malicious.
[DYNAMIC_EXECUTION]: For Library or SDK testing, the skill generates and executes small consumer programs in a scratch directory to test public APIs, as detailed in references/surfaces.md. While necessary for the skill's purpose, this involves the runtime creation and execution of code.
[INDIRECT_PROMPT_INJECTION]: The skill has a high surface area for indirect prompt injection as it is designed to ingest and process data from external, untrusted sources.
Ingestion points: Untrusted data enters the agent context through Web DOM content (clicking, reading), CLI stdout/stderr, and API response bodies (specified in references/surfaces.md).
Boundary markers: The instructions do not define specific delimiters or "ignore embedded instructions" warnings for the content being tested.
Capability inventory: The skill has the capability to execute shell commands (CLI binaries), run Python scripts (prose-lint.py), and write/execute temporary source files (SDK testing).
Sanitization: No escaping, validation, or filtering of the external product output is mentioned before it is processed for report generation.

e2e-test