The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses project environment files such as .env, .env.local, and .env.development to determine the configured application port.
[COMMAND_EXECUTION]: The skill executes several shell commands to interact with version control (git, gh) and to manage local development servers (bin/dev, bin/rails, npm run dev).
[REMOTE_CODE_EXECUTION]: The skill invokes agent-browser skills get core, which retrieves and loads additional operational workflows from an external repository during execution.
[PROMPT_INJECTION]: The skill contains an indirect prompt injection surface as it processes unvalidated repository data, such as file paths and PR metadata, to determine routes and test logic.
Ingestion points: Output from git diff and gh pr view in SKILL.md.
Boundary markers: Not present when mapping changed files to testable routes.
Capability inventory: Full shell command execution and browser automation capabilities across SKILL.md and referenced driver files.
Sanitization: Port extraction utilizes a numeric regex pattern ([0-9]{4,5}), but data from git or PR sources is not sanitized before use.

test-browser