The Agent Skills Directory

[COMMAND_EXECUTION]: Executes git commands to manage local repository state (stash and branch switching) to baseline accessibility audits. It implements security checks on user-provided branch names using git check-ref-format and validation against option-like strings (leading dashes) to mitigate command injection risks.
[EXTERNAL_DOWNLOADS]: Fetches the @accesslint/cli and @accesslint/chrome tools from the public npm registry. These are specialized tools for DOM accessibility auditing.
[DYNAMIC_CONTEXT_INJECTION]: Employs dynamic shell execution during the skill loading phase to resolve the project's default branch. This is used strictly for environment configuration and does not handle untrusted user input.
[REMOTE_CODE_EXECUTION]: Utilizes node -e to execute a small, hardcoded script for parsing the port number from a local service's JSON output. This is a controlled execution environment with no external data interpolation into the script body.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from external URLs which could potentially contain malicious instructions.
Ingestion points: Live web pages accessed via the provided URL in $ARGUMENTS.
Boundary markers: None; the output of the audit is processed directly by the agent.
Capability inventory: Subprocess calls for git and @accesslint/cli found in SKILL.md.
Sanitization: None; the skill relies on the auditing tool to parse the DOM and return violation data.
[SAFE]: No malicious patterns, data exfiltration, or obfuscation techniques were detected. The skill's behavior aligns with its stated purpose of providing accessibility diffs.

accesslint-diff