The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external GitHub pull requests, including titles, bodies, and file content. This creates a surface for indirect prompt injection where malicious instructions in a PR could attempt to bias the agent's classification or review logic.- [SAFE]: All external references and interactions target the toss/es-toolkit repository. As toss is a recognized organization, these references are considered safe. The use of standard development tools like yarn and gh for testing and benchmarking is consistent with the skill's stated purpose.

pr-triage