The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's workflow in SKILL.md instructs the agent to execute the swiftlint command-line tool on "scoped paths" derived from user-supplied diffs or pull requests. This creates a potential command injection vector if the agent environment executes this via a shell without proper sanitization of file names.
[INDIRECT_PROMPT_INJECTION]: The skill is primarily designed to ingest and analyze untrusted data in the form of Swift code and pull requests. 1. Ingestion points: The agent reads user-provided code files and diff segments as specified in SKILL.md. 2. Boundary markers: The instructions lack explicit guidance on how to ignore or handle instructions that might be embedded within code comments or string literals in the analyzed files. 3. Capability inventory: The skill has the capability to execute the swiftlint tool and route tasks to other specialist skills. 4. Sanitization: There are no instructions for sanitizing or filtering the content of the analyzed files before processing.
[EXTERNAL_DOWNLOADS]: The skill provides links to external documentation on sosumi.ai across multiple reference files such as references/concurrency.md and references/observation.md. While these appear to be documentation mirrors, they originate from a non-standard domain that is not recognized as a trusted organization or well-known service.

swift-code-review