The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on shell command execution via the subprocess.run function in several scripts, including scripts/gh_git.py, scripts/task_scaffold.py, and scripts/test_cross_agent_coordination.py. These commands are used to interface with the git CLI for essential repository operations like cloning, branching, committing, and pushing. The implementation uses list-based arguments which is a standard safety practice against shell injection.
[EXTERNAL_DOWNLOADS]: The skill performs network operations to interact with the GitHub API (api.github.com) and to clone or fetch data from GitHub repositories. These operations are directed at well-known developer services and are necessary for the skill's cross-agent synchronization features.
[DATA_EXFILTRATION]: Through scripts/email_trigger.py, the skill can package project context, task objectives, and acceptance criteria into email drafts intended for external agent platforms (such as Manus, AnyGen, or Coze). While this involves sending data outside the immediate environment, it is the explicitly stated primary function of the skill to enable cross-platform task delegation.
[PROMPT_INJECTION]: The skill parses external data from project-tasks.md and Markdown frontmatter. This creates a surface for indirect prompt injection where malicious instructions could be embedded in task descriptions. However, the skill treats this data as informational context and does not execute it as code, and it provides instructions for agents to follow specific workflow boundaries.

cross-agent-coordination