CLI Reference
The PRAXIS CLI is the primary interface for the Truth Kernel. It handles task definition, evidence collection, verification, repair, and reporting.
Global Flags
| Flag | Description |
|---|---|
--help | Show help for any command |
--version | Show PRAXIS version |
Commands
praxis init
Initialize a PRAXIS workspace in the current directory.
praxis init
Creates .praxis/task.yaml (skeleton) and .praxis/runs/ and .praxis/reports/ directories. Safe to run multiple times — will not overwrite existing task specs.
praxis spec
Create or update a task specification.
praxis spec --description "Add a health check endpoint"
praxis spec --file ./custom-task.yaml
Options:
| Flag | Description |
|---|---|
--description <text> | Task description (generates skeleton criteria) |
--file <path> | Load spec from a custom YAML file |
The generated TaskSpec requires human approval (human_approved: true) before verification will proceed.
praxis verify
Run the three verification gates against the current task.
praxis verify
The kernel collects evidence (git diff, command logs, test output) then runs EvidenceGate, ExecGate, and FinalGate. Output is a structured verdict with per-gate results.
Verdict values:
- PASS — All criteria met
- HOLD — Some criteria need attention (evidence gaps, execution gaps)
- FAIL — Task not complete
Options:
| Flag | Description |
|---|---|
--gate <name> | Run a specific gate only (evidence, exec, final) |
--verbose | Show detailed evidence for each gate |
--json | Output verdict as JSON |
praxis repair
Generate a repair packet targeting failed verification criteria.
praxis repair
Outputs a structured RepairPacket listing each failed criterion, the issue detected, and a hint for fixing. Designed to be passed to a coding agent.
Must be run after a failed verify.
praxis status
Show the current state of the PRAXIS workspace.
praxis status
Displays:
- Current task ID and description
- Whether the TaskSpec is human-approved
- Last verification verdict
- Number of runs and reports
Options:
| Flag | Description |
|---|---|
--json | Output as JSON |
praxis report
Generate a signed audit report for the last verification run.
praxis report
Writes a markdown report to .praxis/reports/<run-id>.md containing:
- Task ID and description
- Per-gate results with evidence summaries
- Overall verdict
- Timestamp and run ID
Exit Codes
| Code | Meaning |
|---|---|
0 | PASS — all gates passed |
1 | FAIL — one or more gates failed |
2 | HOLD — gates passed with gaps |
3 | Error — execution failed (invalid config, missing task, etc.) |
Environment Variables
| Variable | Description | Default |
|---|---|---|
PRAXIS_HOME | PRAXIS data directory | .praxis/ |
PRAXIS_LOG_LEVEL | Log level | info |
PRAXIS_NO_COLOR | Disable colored output | false |