The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions involve analyzing external inputs such as changed code behavior, interface definitions, and failure logs to design and execute tests. This introduces a potential surface for indirect prompt injection where malicious instructions embedded in the analyzed code or data could attempt to influence the agent. However, the skill explicitly advocates for safety-conscious practices, such as using sanitizers during fuzzing and performing experiments in isolated staging environments first.
[COMMAND_EXECUTION]: The skill provides numerous examples for executing standard testing and build tools (e.g., pytest, npm test, k6 run, afl-fuzz, afl-gcc). These operations are necessary for the skill's primary purpose of verification and quality engineering. The instructions emphasize bounded analysis and the use of existing, installed tools rather than arbitrary remote execution.
[EXTERNAL_DOWNLOADS]: The skill references various well-known developer services and testing tools (e.g., Stripe, Pactflow, LaunchDarkly, Percy, Chromatic, Gremlin). It provides examples for installing these tools via official package managers (npm, pip, brew, apt). All external references target established, reputable platforms and organizations for legitimate testing and observability purposes.

using-quality-engineering