The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/github-issue-handler.rb script is vulnerable to shell command injection. It constructs commands for the gh CLI by joining unescaped strings (such as repository names and issue references) and executing them via backticks in the execute_gh method. An attacker who can influence the issue reference or repository name (e.g., through a malicious URL or CLI input) could execute arbitrary shell commands on the host system.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8). It fetches content from GitHub issues and parses it into acceptance criteria using scripts/acceptance-criteria.rb. These criteria are then used to drive the logic of the iterative debug loop. Malicious instructions embedded within a GitHub issue could influence the agent's behavior during debugging.\n
Ingestion points: Issue bodies fetched from GitHub in scripts/github-issue-handler.rb.\n
Boundary markers: None; the parser relies on Markdown headers to identify sections.\n
Capability inventory: The skill can execute shell commands, perform file system operations, and control a web browser.\n
Sanitization: No sanitization or escaping is performed on the external Markdown content before it is processed by the debug state machine.\n- [CREDENTIALS_UNSAFE]: The skill stores usernames and passwords in plaintext in the .frontend-debug/config.json file. Although scripts/config-manager.rb attempts to protect this file using .gitignore and local file permissions (0600), plaintext storage of credentials remains a significant security risk if the local filesystem is accessed by an unauthorized party.

frontend-debug