The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized operations were detected. The skill is designed for read-only analysis of project files and writing only to a specific session state file.
[PROMPT_INJECTION]: The skill ingests untrusted data by reading implementation files and documentation, which constitutes an indirect prompt injection surface. However, the risk is negligible due to the absence of dangerous tools.
Ingestion points: implementation files in the files_changed array and project documentation files (SKILL.md, Steps 1.2 and 1.3).
Boundary markers: No explicit delimiters or markers are used to separate file content from the agent's internal instructions.
Capability inventory: The skill is restricted to reading local files and writing only to the .ws-session/verifier.json session file. It has no access to the network, shell, or arbitrary code execution (eval/exec).
Sanitization: There is no mention of sanitizing or escaping the content of the files being reviewed.

ws-verifier