The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data during its review process, creating a vulnerability surface for indirect instructions embedded in code or diffs.
Ingestion points: The workflow explicitly requires ingesting 变更文件 (changed files), diff, and 已有定向测试/集成证据 (existing test/integration evidence) from the workspace (SKILL.md).
Boundary markers: There are no instructions provided to use specific delimiters or to warn the agent to ignore instructions embedded within the reviewed code or diff data.
Capability inventory: The skill has the capability to trigger project-native checks, targeted tests, integration checks, and call upon other agents/models (omo-adaptive-execution, Oracle) to process the ingested data.
Sanitization: The skill lacks instructions for sanitizing or escaping the content of the files and diffs before they are presented to the reviewer agents.

review-work