The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, potentially untrusted code diffs and source files. It instructs the agent to use tools like git diff, sed, and nl to ingest this data into its context. The risk is mitigated by explicit instructions to perform a semantic pass before running any code and by limiting execution to reproduction of identified findings only.- [COMMAND_EXECUTION]: The skill employs standard command-line utilities for code inspection, including git, sed, and nl. These tools are used for their intended purpose of reading and comparing files in a development environment. The instructions specifically caution against running broad test suites or build commands initially, promoting a principle of least privilege in execution.

defect-first-review