The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the user to install the orbit CLI using a high-risk pattern: curl -sSfL https://raw.githubusercontent.com/jorgemuza/orbit/main/install.sh | sh. This executes a remote shell script directly with shell privileges.
[EXTERNAL_DOWNLOADS]: Dependencies and tools are fetched from non-whitelisted external sources, including a personal GitHub repository (jorgemuza/orbit) and a custom Homebrew tap (jorgemuza/tap/orbit).
[COMMAND_EXECUTION]: The skill heavily relies on executing shell commands through the orbit CLI to interact with local databases and session files.
[DATA_EXFILTRATION]: The skill includes functionality to push token usage data to a centralized analytics service called 'Draxarp' (orbit -p paybook tracking tkm push), which involves sending data from the local environment to an external network endpoint.
[PROMPT_INJECTION]: The skill exhibits an Indirect Prompt Injection surface (Category 8) as it ingests untrusted data from Claude Code session files (~/.claude/projects/**/*.jsonl) and uses that data to perform calculations and shell-based operations.
Ingestion points: Reads JSONL files from the user's local .claude directory (SKILL.md).
Boundary markers: No explicit boundary markers or 'ignore embedded instructions' warnings are present when processing session data.
Capability inventory: Uses orbit subprocess calls for syncing, tracking, and pushing data.
Sanitization: No evidence of sanitization or validation of the session file content before processing.

tkm