The Agent Skills Directory

[DATA_EXFILTRATION]: The skill instructions in concurrency-patterns/SKILL.md and concurrency/SKILL.md direct the agent to read a specific file from the user's Downloads directory (~/Downloads/docs/Swift-Concurrency-Updates.md). Accessing files outside the project workspace can lead to unintended data exposure.
[INDIRECT_PROMPT_INJECTION]: By encouraging the agent to read external documentation from a sensitive user directory (~/Downloads), the skill introduces an attack surface for indirect prompt injection. If a malicious file with that name is placed in the Downloads folder, it could influence the agent's behavior.
Ingestion points: The agent is instructed to read ~/Downloads/docs/Swift-Concurrency-Updates.md if present.
Boundary markers: The instructions lack delimiters or warnings to ignore potentially embedded instructions in the external file.
Capability inventory: The skill utilizes high-privilege tools including Bash, Write, and Edit across its modules.
Sanitization: There are no instructions for validating or sanitizing the content of the external documentation before processing.
[COMMAND_EXECUTION]: The code-size/SKILL.md module leverages the Bash tool to execute swiftlint for tasks such as creating violation baselines and calculating error counts. While these are standard development tasks, the use of Bash represents a high-capability execution environment.

swift-development