The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the subprocess module across several files (scripts/accept_changes.py, scripts/office/soffice.py, scripts/office/validators/redlining.py) to execute system utilities including LibreOffice (soffice), git, and gcc.\n- [REMOTE_CODE_EXECUTION]: The script scripts/office/soffice.py performs runtime compilation and process injection. It writes a C source string to a temporary file in /tmp, compiles it into a shared library using the system gcc compiler, and then injects this library into the soffice process using the LD_PRELOAD environment variable. While documented as a compatibility shim for sandboxed environments, these techniques are highly characteristic of sophisticated malware.\n- [PROMPT_INJECTION]: The skill frontmatter in SKILL.md and the LICENSE.txt file explicitly claim the author is 'anthropic' and that the skill is proprietary to 'Anthropic, PBC'. This information is deceptive as it contradicts the 'ChatAndBuild' author context provided for this analysis, representing a metadata-based impersonation of a trusted organization to deceive users regarding the skill's origin and safety.

docx