The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's primary function is to launch and manage parallel agent processes via tmux and shell scripts. It dynamically generates .run.sh script files for each worker and executes them in separate tmux windows. This behavior is the central, documented purpose of the skill.
[EXTERNAL_DOWNLOADS]: The skill documentation specifies prerequisites such as bash, tmux, jq, and flock, instructing the user to install them via standard system package managers. This is a common and safe practice for tools with complex dependencies.
[SAFE]: The skill implements defensive programming, such as the validate_fleet_id function in lib/logging.sh, which uses regular expressions to prevent shell injection attacks in fleet and worker identifiers. This demonstrates a high level of security awareness from the author.
[SAFE]: Analyzed the attack surface for indirect prompt injection.
Ingestion points: Worker logs (session.jsonl) and status files are ingested by monitoring scripts (feed.sh, report.sh, status.sh, view.sh).
Boundary markers: Data is processed in structured JSONL format.
Capability inventory: Monitoring scripts perform read and display operations; orchestrator scripts manage agent processes in tmux.
Sanitization: All log ingestion uses robust parsers like jq and Python's json.loads(), which prevents untrusted log content from influencing the orchestrator's execution logic.

dag-fleet