dark-factory-commands
Common commands and CLI invocation reference for the Dark Factory DOT pipeline runner — running smoke and full gated pipelines, CXDB failure diagnosis with df-healer, pipeline graph visualization, and
SKILL.md.claude/skills/dark-factory-commands/SKILL.md
GitHubDark Factory Common Commands
Use this skill when you need the standard command syntax for running Dark Factory pipelines, diagnosing CXDB execution logs with the Healer, rendering DOT pipeline graphs, or executing the test suite.
When to use
- Running a smoke or sanity pipeline without LLM calls.
- Running a full gated pipeline with CXDB recording and Agent Orchestrator (AO).
- Clustering CXDB execution failures into a Healer diagnosis.
- Rendering
.dotpipeline graphs to PNG for visual inspection. - Running pytest against the test suite, specific test files, or individual test cases.
Pipeline execution
Smoke pipeline
Echo backend, no LLM calls:
dark-factory --pipeline pipelines/factory/hello.dot --goal "smoke test" --backend echo
Full gated pipeline
Run the full gated pipeline with CXDB recording from the target repo cwd:
dark-factory \
--pipeline pipelines/factory/gates.dot \
--goal "<feature description>" \
--backend ao \
--ao-agent antigravity \
--feature <feature_name> \
--cxdb ~/.dark-factory/cxdb.sqlite
gates.dot declares hello as its default holdout feature; an explicit --feature value overrides that DOT default at runtime.
Diagnosis and visualization
Cluster CXDB failures into a Healer diagnosis
df-healer --cxdb ~/.dark-factory/cxdb.sqlite
Visualize a pipeline graph
dot -Tpng pipelines/factory/gates.dot -o gates.png
Testing
Run the full suite, a single file, or an individual test:
.venv/bin/python -m pytest tests/
.venv/bin/python -m pytest tests/test_engine.py -k green
.venv/bin/python -m pytest tests/test_gates.py::test_parse_verdict_pass_warn_fail
Legacy dev-only:
.venv/bin/python -m runner ...from$DARK_FACTORY_HOME.