The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill identifies and reads project configuration files (e.g., vitest.config.ts, pyproject.toml, Cargo.toml) and source code to determine the appropriate testing framework and strategy. This ingestion of untrusted repository data presents an indirect prompt injection attack surface where malicious instructions in these files could influence agent behavior.
Ingestion points: Project configuration files (package.json, pyproject.toml, etc.) and test files (*_test.go, tests/*.py).
Boundary markers: Absent; the skill does not specify delimiters or instructions to ignore embedded prompts in processed files.
Capability inventory: Execution of test runners (pytest, vitest, cargo test, go test), shell command execution (command -v k6), and file system access.
Sanitization: Absent; no explicit validation or escaping of external content is described before processing.
[EXTERNAL_DOWNLOADS]: The skill suggests installing various external packages and tools via standard package managers to enable specific features such as accessibility scanning, parallelization, and load testing.
Evidence: Instructions include npm install -D @axe-core/playwright, pip install pytest-xdist, cargo install cargo-nextest, and npm init playwright@latest. It also references the official Playwright Docker image (mcr.microsoft.com/playwright). These references target well-known package registries and official repositories.
[COMMAND_EXECUTION]: The skill relies on executing shell commands to run test suites, perform benchmarks, and check for the existence of required tools.
Evidence: Shell patterns detected include command -v k6, go test -parallel N, pytest -n auto, and npx playwright test. These are standard development operations within the scope of the skill's purpose.

testing