The Agent Skills Directory

Indirect Prompt Injection (HIGH): The skill processes untrusted external content (issue reports and bug descriptions) while possessing capabilities to modify files and execute commands. This creates a risk where malicious instructions in a bug report could lead to unauthorized code changes.
Ingestion points: .issues/opening.md and user-provided bug reports.
Capability inventory: Modifies source files and executes shell commands (ruff, pytest, npm build).
Boundary markers: No delimiters or explicit warnings to ignore embedded instructions are present.
Sanitization: No evidence of input validation or sanitization.
Command Execution (MEDIUM): The skill executes local build and test commands (python -m pyright, ruff, pytest, npm run build). While these are standard development tools, they are executed on a codebase that the skill is actively modifying based on potentially untrusted external inputs.
Remote Code Execution (LOW): While no direct remote downloads are seen, the skill's reliance on npm run build and npm test can trigger remote package fetching or execution if the package.json is modified.

issue-fixer