CLI Reference
Every non-TUI entry point in one place. Run wallbreaker --help or wallbreaker <command> --help for the authoritative list.
One-Shot & TUI
| Invocation | What it does |
|---|---|
wallbreaker | Launch the TUI |
wallbreaker "prompt" | One-shot prompt, then exit |
wallbreaker --auto "prompt" | Run autonomously until finish()/ask_operator() |
wallbreaker --auto --rounds 5 "prompt" | Cap autonomous rounds (default 12) |
wallbreaker --resume | Resume the autosaved session (sessions/autosave.json) |
wallbreaker --resume mysession.json | Resume a specific session file |
Global endpoint flags (available on the main command): --config, --profile, --base-url, --model, --protocol, --api-key-env, --api-key, plus --system (system prompt override), --no-tools, and target overrides --target, --target-model, --target-modality {text,image}.
Library & Transform Commands
| Command | What it does |
|---|---|
wallbreaker lib update|list|path | Manage the L1B3RT4S jailbreak library |
wallbreaker parsel update|list|path | Manage the P4RS3LT0NGV3 transform library (MCP server backend; needs Node.js on PATH) |
wallbreaker eni list|update|path | Browse the ENI persona-jailbreak collection |
wallbreaker transform "leet,base64" "text" | Apply a transform chain on the CLI (--decode to reverse) |
Run Logs & Reporting
All log-taking commands accept a run log path or a directory (default: latest in sessions/):
| Command | What it does |
|---|---|
wallbreaker findings [log] | List bypasses (COMPLIED/PARTIAL) from a run log |
wallbreaker report [log] | Render a findings report (markdown to stdout) |
wallbreaker report [log] --html --out report.html | Styled HTML report to a file |
wallbreaker export [log] --out findings.json | Structured findings JSON |
wallbreaker export [log] --fail-on-finding | CI gate: exit code 2 if any bypass present |
wallbreaker regrade [log] --config config.toml | Re-judge a run log with the current judge |
InfoThe dashboard can serve any run-log directory:
wallbreaker dashboard --sessions sessions/ (default). See WebUI V2.CI Gates
| Command | What it does |
|---|---|
wallbreaker export [log] --fail-on-finding | Fails (exit 2) if any bypass exists in the log |
wallbreaker baseline save [log] --out baseline.json | Write an ASR baseline from a run log |
wallbreaker baseline compare [log] --baseline baseline.json | Fail (exit 2) if ASR regressed beyond --max-regression (default 0.05 per technique) |
Configuration & Dashboard
| Command | What it does |
|---|---|
wallbreaker check [--config config.toml] | Validate config and print a readiness checklist |
wallbreaker dashboard | Serve the web dashboard on 127.0.0.1:8787 |
wallbreaker dashboard --host 0.0.0.0 --allow-network | Expose beyond loopback (explicit risk acknowledgement) |
wallbreaker dashboard --sessions sessions/ | Point the dashboard at a run-log directory |
InfoCorpus integrity (the
library.lock.toml SHA pins) is enforced automatically at load time — a mismatched runtime-fetched corpus fails closed. The interactive corpus verify [--update] command (also aliased as parsel verify) resolves actual HEAD SHAs via git ls-remote and updates the lock file. It shipped in roadmap-implementation (TG3) but its CLI dispatch was lost in the PR #21 × PR #24 merge; it is being restored in PR #34 .Last updated on