The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/export_context.py uses subprocess.run to execute git commands and the pbcopy utility. This allows the skill to interact with the local operating system and clipboard functionality.\n- [DATA_EXFILTRATION]: The skill collects and prepares project-sensitive data for export. The read_files function in scripts/export_context.py lacks strict path validation, creating a potential directory traversal vulnerability where an agent could be manipulated into reading files outside the intended project directory by using relative paths like ../../.\n- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through the data it aggregates.\n
Ingestion points: In scripts/export_context.py, the script reads git commit history, diffs, and arbitrary file contents provided via the --include-files flag.\n
Boundary markers: The output briefing document uses markdown headers and code blocks to separate sections, but lacks explicit instructions to ignore any agent-directives found within the ingested data.\n
Capability inventory: The script uses subprocess.run to call git and pbcopy in scripts/export_context.py.\n
Sanitization: No sanitization or escaping is applied to the content read from files or git, allowing malicious instructions in the codebase to propagate into the exported document.

cowork-export