The Agent Skills Directory

[SAFE]: No security vulnerabilities or malicious patterns were identified in the skill. It is designed for linguistic processing and follows clear procedural rules.
[COMMAND_EXECUTION]: The skill includes several TypeScript utility scripts (e.g., manifest-sync.ts, scaffold-eval-loop.ts) designed to run in the Bun environment. These scripts perform local file operations such as creating directories and writing JSON/TSV metadata. They include robust security checks, such as using realpathSync to prevent path traversal and explicitly refusing to follow symbolic links (lstatSync(path).isSymbolicLink()) to protect against link-based attacks.
[PROMPT_INJECTION]: The agent instruction files focus on register calibration and translation artifact detection. They do not contain any instructions that attempt to bypass safety guidelines or override the agent's core programming.
[DATA_EXFILTRATION]: All file operations are restricted to the local project directories (docs/forsvn and .forsvn). There are no network requests to untrusted external domains for data exfiltration.
[SAFE]: The skill process involves analyzing user-provided Vietnamese text. It mitigates potential indirect prompt injection by enforcing strict meaning preservation rules and using a critic agent to verify the output against the original source.

polish-vn