The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands such as git, gh, npx, and uv. The command gh pr diff <n> utilizes a placeholder for a pull request number; if the agent accepts this value from an untrusted source without validation, it could be exploited for command injection.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from the project's source code and version control diffs, which could contain malicious instructions.
Ingestion points: Frontend source files in src/frontend/** and the output of git diff or gh pr diff commands.
Boundary markers: Absent. The instructions do not specify delimiters or warnings to ignore embedded content in the files being processed.
Capability inventory: The skill possesses the ability to modify source files (to apply accessibility fixes) and execute shell commands for testing.
Sanitization: Absent. There is no explicit logic to sanitize or filter the content of the files before they influence the agent's routing or remediation logic.
[EXTERNAL_DOWNLOADS]: The skill installs Playwright and its browser dependencies using uv run --with playwright playwright install chromium. This is a routine operation for automated browser testing and utilizes a well-known service.

ibm-a11y-pr-remediation