The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions in AGENTS.md direct the agent to perform shell-based operations using 'tsc --noEmit' via various package managers (npm, pnpm, bun, yarn) to verify the type correctness of test files.- [INDIRECT_PROMPT_INJECTION]: The skill operates on untrusted test files (.test.ts) provided in the project context. While it instructs the agent to use standard tools like the TypeScript compiler, the combination of code ingestion and command execution creates a vulnerability surface. 1. Ingestion points: Test files (.test.ts) and project configuration files identified for review. 2. Boundary markers: No explicit markers or instructions to ignore embedded commands are present in the guidelines. 3. Capability inventory: Shell command execution (tsc) and package management (npm) as described in AGENTS.md and SKILL.md. 4. Sanitization: The instructions do not specify sanitization or escaping of file paths or content before they are used in shell commands.- [EXTERNAL_DOWNLOADS]: The skill documentation suggests the installation of well-known and trusted development packages from the NPM registry, including 'vitest', '@vitest/coverage-v8', and 'fast-check'. These are standard tools in the JavaScript ecosystem.- [PROMPT_INJECTION]: The skill uses highly imperative language ('MANDATORY', 'CRITICAL REMINDER', 'NON-NEGOTIABLE') in AGENTS.md and SKILL.md to constrain the agent's behavior, specifically forcing the execution of type-checking workflows before a task can be considered complete.

accelint-ts-testing