The Agent Skills Directory

[COMMAND_EXECUTION]: The main script scripts/launch_check.py utilizes subprocess.run to interface with the git command-line tool for identifying project files and to execute bundled audit scripts (geo_audit.py and polish_audit.py). These calls are implemented using list-based arguments rather than shell strings, adhering to secure coding practices for process execution.\n- [EXTERNAL_DOWNLOADS]: To perform mobile design and browser-based audits, the skill requires the playwright library. It provides instructions to install the package and its browser binaries via pip, including a requirement that the agent must obtain explicit user confirmation before initiating the installation.\n- [INDIRECT_PROMPT_INJECTION]: The skill analyzes external websites and local repository content, creating an ingestion surface for untrusted data in scripts/launch_check.py, scripts/geo_audit.py, and scripts/polish_audit.py. Maliciously crafted content in these sources could attempt to influence the agent's output. The skill mitigates this risk by instructing the agent to treat findings as untrusted evidence and by employing sanitization routines like sanitize_report to redact secrets and escape HTML in reports.\n- [OBFUSCATION]: A regular expression in scripts/launch_check.py contains Unicode directional control characters (U+202A through U+202E and U+2066 through U+2069). This is a defensive measure used by the clean function to identify and remove hidden control characters from audited text, protecting the user's terminal from potential injection or display manipulation attacks.

launch-check