The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the dsh CLI to manage terminal blocks and tabs. The scripts/danteterm_delegate.py script uses subprocess.run to execute these commands, which is necessary for the skill's primary functionality within the DanteTerm environment.
[INDIRECT_PROMPT_INJECTION]: The delegation feature involves passing prompts to sub-agents and reading their output. The script facilitates this by reading scrollback or redirected result files.
Ingestion points: Data enters the context via dsh termscrollback output or reading temporary result files in scripts/danteterm_delegate.py.
Boundary markers: The skill does not implement specific delimiters for external agent output.
Capability inventory: The skill can execute commands via dsh, create/delete blocks, and manage files via dsh file.
Sanitization: The script uses shlex.quote to sanitize prompts passed to sub-agents, mitigating command injection into the shell environment of the delegated block.
[DYNAMIC_EXECUTION]: The delegation script dynamically constructs shell commands to redirect agent output to temporary files and signal completion. These commands are assembled using fixed templates and sanitized user input, representing a standard orchestration pattern for terminal-based automation.

danteterm