The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's primary functionality is implemented by executing the hledger binary and internal Python scripts (scripts/tab_calc.py, scripts/tab_helpers.py) using subprocess.run to interact with ledger files based on user input.\n- [EXTERNAL_DOWNLOADS]: The references/installing-hledger.md documentation provides links to fetch the hledger binary from its official project repository on GitHub.\n- [REMOTE_CODE_EXECUTION]: The installation guide suggests a command that downloads and extracts a binary archive from the hledger project's GitHub releases directly to system paths using a piped command.\n- [COMMAND_EXECUTION]: The agent is instructed to run mise install automatically and without user confirmation if a configuration file is detected in the project root, which could allow for the execution of installation logic defined in local project files.\n- [COMMAND_EXECUTION]: The installation reference suggests using sudo to acquire administrative privileges when installing prerequisites via system package managers like apt or xbps.\n- [PROMPT_INJECTION]: The skill processes untrusted data from user descriptions and the notes field in tab.yaml, creating an indirect prompt injection surface when generating summaries or journal entries.\n
Ingestion points: User-provided expense descriptions and the notes array in tab.yaml.\n
Boundary markers: None provided in prompt instructions or script outputs.\n
Capability inventory: Execution of shell commands and Python scripts using subprocess.run.\n
Sanitization: No validation or escaping is performed on external text inputs before processing.

split-expenses