Skip to main content
The bundled helper script scripts/task_loop.py provides three CLI subcommands for managing repo-local task artifacts. Run it from inside the target repository using the path to wherever you installed the skill.
Requires Python 3.10 or later.

​
Setting SKILL_PATH

Set SKILL_PATH to the installed skill directory before running any commands. The path depends on which tool you are using. 
SKILL_PATH=.agents/skills/repo-task-proof-loop
All examples in this reference use $SKILL_PATH/scripts/task_loop.py as the script path. Run commands from inside the target repository.

​
Available subcommands

init

Create .agent/tasks/<TASK_ID>/, install project-scoped subagents, and update guide files.

validate

Validate required task files and JSON structures. Exits 1 if anything is missing or malformed.

status

Summarize current task artifact status, including evidence and verdict overall status.

​
Workflow phases are not CLI subcommands

The workflow phases freeze, build, evidence, verify, fix, and run are skill-level commands directed at the agent. They are not subcommands of task_loop.py. The CLI only provides init, validate, and status.
Always run init to completion before any workflow phase. Never overlap init with freeze, build, evidence, verify, fix, or subagent spawning.

​
Quick reference

# Initialize a new task
python3 "$SKILL_PATH/scripts/task_loop.py" init --task-id <TASK_ID>

# Validate a task artifact set
python3 "$SKILL_PATH/scripts/task_loop.py" validate --task-id <TASK_ID>

# Summarize task status
python3 "$SKILL_PATH/scripts/task_loop.py" status --task-id <TASK_ID>

Build docs developers (and LLMs) love

Get started for freeTalk to us