The missing process layer for specification-driven development.
Drydock gives specification-driven development the part it skipped: a repeatable Agile delivery and Test Driven Development based end-to-end process that turns specs into reviewed plans, context-managed builds, and obvious ways to maintain.
Install · 60-Second Example · Why It Is Different · Canonical Specification Overview Deck · White Paper · 10 Minute Overview · User Installation Guide
uv tool install drydock-sdd # then: drydock init MyApp && drydock status
Drydock runs on your existing Claude or Codex subscription CLI. It does not require API keys or per-token API billing.
Drydock adds the missing process layer to specification-driven development: import source material, analyze it into stories and acceptance criteria, review decisions in the QuarterDeck, plan a dependency graph, build one context-optimized step at a time, verify evidence, and keep the Blueprint and software aligned as the product changes.
You are the Commander— the product owner. The LLM is your Agile delivery team. Drydock makes that relationship explicit, reviewable, and repeatable.
Copyright (c) 2026 Web Cloud Studio. Licensed under the MIT License. See LICENSE.
uv tool install drydock-sdd
drydock config set llm_provider claude
drydock init MyApp --description "A small web application."
drydock import MyApp ./notes --format markdown
drydock analyze MyApp
drydock run quarterdeck MyApp
drydock plan MyApp
drydock run quarterdeck MyApp
drydock build MyApp
drydock refit MyApp
drydock build MyApp
That loop creates a Target workspace, decomposes the source material, opens the Commander review surface, builds a dependency graph, and executes the first runnable frontier with persisted evidence.
Drydock is an installable Python command-line package. The PyPI distribution is
drydock-sdd
; the installed command is drydock
.
Drydock implements the SAIL methodology:
| Phase | Purpose | Primary commands |
|---|---|---|
| Set Up | Install, configure, and initialize a Target workspace | config , init , status |
| Analyze | Import material and decompose it into stories, blockers, and acceptance milestones | import , analyze , run quarterdeck , plan |
| Implement | Build the Manifest frontier and verify evidence | build , build status , build verify , rigging , document |
| Loop | Manage change while preserving the Blueprint as source of truth | refit , build , document |
The core idea is simple: reproducible LLM builds require a process. Drydock uses Agile structure, explicit product-owner review, durable evidence, and context-managed build prompts so generated software can be inspected, repeated, and iterated.
Drydock is for subscription-authenticated CLI users.
It does not use API-key-backed model calls and does not require per-token API billing. LLM-assisted commands execute through a locally authenticated provider CLI:
claude
for Anthropic Claude subscription CLI users.codex
for OpenAI Codex subscription CLI users.
Set the provider with:
drydock config set llm_provider claude
drydock config set llm_provider codex
The provider CLI must already be installed, authenticated, and available on PATH
.
Deterministic commands such as status
, validate
, document assemble
, and publish
do not call an LLM.
Python 3.11 or later is required.
Recommended:
uv tool install drydock-sdd
Alternative:
pipx install drydock-sdd
Virtual environment install:
python -m pip install drydock-sdd
Verify:
drydock --version
drydock --help
PDF publishing (drydock publish --pdf
) is optional and requires the pdf
extra plus a local Chromium download:
uv tool install "drydock-sdd[pdf]"
playwright install chromium
See docs/USER_INSTALLATION.md for the full installation guide.
Drydock ships two Claude Code skills that support the Loop phase: /refit
(design
discussion captured to a notes file in the Target directory) and /apply-refit
(turn the captured decisions into change tickets). Installing them is optional:
python -c "import shutil, pathlib; from drydock.paths import get_rigging_root; \
dest = pathlib.Path.home() / '.claude' / 'skills'; \
[shutil.copytree(s, dest / s.name, dirs_exist_ok=True) \
for s in (get_rigging_root() / 'skills').iterdir()]"
The skills are copied from the packaged Rigging/skills/
directory into
~/.claude/skills/
. See docs/Drydock_SKILLS.md for usage.
Create one projects directory, configure Drydock's workspace and build output root, and initialize a Target.
export PROJECTS="$HOME/projects"
mkdir -p "$PROJECTS/drydock"
drydock config set drydock_workspace "$PROJECTS/drydock"
drydock config set drydock_build_directory "$PROJECTS"
drydock config set llm_provider claude
drydock init MyApp --display-name "My App" --description "A working software product."
drydock status
Import source material and run the planning loop:
drydock import MyApp ./notes --format markdown
drydock analyze MyApp
drydock run quarterdeck MyApp
drydock plan MyApp
drydock build status MyApp
Build and verify one frontier at a time:
drydock build MyApp
drydock build status MyApp
drydock build verify MyApp <step-id>
The Target workspace lives under:
$DRYDOCK_WORKSPACE/targets/<Target>/
The generated application is written under:
$DRYDOCK_BUILD_DIRECTORY/<Target>/
Drydock is not a prompt collection and it is not a one-shot code generator. It is a delivery system with durable artifacts:
- Blueprint: typed Markdown specifications that remain the source of truth.
- Manifest: the executable dependency graph for build order, dependencies, and state.
- QuarterDeck: the web review surface where the product owner answers questions, reviews stories, and directs the process.
- Compass files: persistent product-owner intent injected into the right command runs.
- Soundings: acceptance checklist and implementation evidence.
- Sea Trials: product-level objectives and proof-of-delivery criteria.
- Rigging: shared branding, stack rules, templates, and compact context derivatives.
- Execution logs: reproducible prompt, raw output, stderr, event, and result artifacts.
The Commander is the product owner. The LLM is treated as an Agile delivery team. Drydock's job is to make that relationship explicit, reviewable, and repeatable.
Drydock 0.1.2
is an alpha release. The primary SAIL path is implemented, but
the command surface and Typed Specification contracts remain unstable during
the 0.x
series:
- Workspace configuration and Target initialization.
- Markdown, source tree, Spec Kit, and Compass import.
- LLM-assisted analysis with blockers, questionnaires, Soundings, Sea Trials, and Commander review artifacts.
- LLM-assisted planning into typed Blueprint files and
MANIFEST.md
. - Manifest-frontier build execution, evidence capture, and human verification.
- Refit change-ticket conformance and applied-spec drift reconciliation.
- QuarterDeck runtime for review and process navigation.
- Rigging compaction, update, and verification.
- Target documentation generation and assembly.
- Deterministic Markdown publishing to HTML and optional PDF.
- Survey and prompt-review tooling used to evaluate Drydock's own command quality.
One command surface remains visible but deferred in the current CLI:
drydock build score <Target>
The installed wheel includes Drydock's canonical product specification as a
read-only package resource at drydock/resources/docs/Drydock_Specification.md
.
Read docs/SOUNDINGS.md for the authoritative implementation readiness checklist and test evidence.
drydock --help
drydock --version
drydock config show
drydock config set <key> <value>
drydock init <Target> [--display-name <name>] [--description <desc>]
drydock status [<Target>]
drydock validate <Target> [--verbose]
drydock run quarterdeck [<Target>] [--host HOST] [--port PORT]
drydock import <Target> <Source> --format <auto|markdown|source|speckit|compass>
drydock analyze <Target> [--model <model>] [--llm-provider <claude|codex>]
drydock plan <Target> [--model <model>] [--llm-provider <claude|codex>]
drydock build <Target> [--step <step-id>] [--force] [--build-dir <path>]
drydock build status <Target>
drydock build verify <Target> <step-id>
drydock build score <Target>
drydock rigging compact [<Target>] [--all] [--force]
drydock rigging compact <Target> --include-file <file.md>
drydock rigging update <Target> [--dry-run]
drydock rigging verify <Target>
drydock document generate <Target> [--model <model>]
drydock document assemble <Target> [--theme <theme>]
drydock document <Target> [--model <model>] [--theme <theme>]
drydock publish <Source.md> --output <Output.html> [--theme <theme>] [--flatten] [--pdf]
drydock survey <Target> [--run] [--import <path>] [--command <name>] [--raw]
drydock prompt review <component>
drydock shipslog [--dir <path>] [--dry-run]
Configuration keys:
| Key | Environment override | Purpose |
|---|---|---|
drydock_workspace |
||
DRYDOCK_WORKSPACE |
||
Workspace containing targets/ and Drydock logs |
||
drydock_build_directory |
||
DRYDOCK_BUILD_DIRECTORY |
||
| Root where generated applications are written | ||
drydock_model |
||
DRYDOCK_MODEL |
||
| Default model for LLM-assisted commands | ||
llm_provider |
||
LLM_PROVIDER |
||
Subscription CLI provider: claude or codex |
||
prompt_warn_tokens |
||
PROMPT_WARN_TOKENS |
||
| Prompt-size warning threshold in tokens | ||
quarterdeck_port |
||
QUARTERDECK_PORT |
||
| Default QuarterDeck port | ||
shipslog_dir |
||
DRYDOCK_SHIPSLOG_DIR |
||
Ship's Log posts package directory for drydock shipslog |
Public hub and launch materials:
Web Cloud StudioDrydock GitHub repositoryCanonical Drydock specificationLaunch deck and presentationLaunch videoImproving Step Accuracy in Specification-Driven Development
Repository references:
Install DrydockDrydock specification sourceRendered specification HTMLRendered specification PDFLaunch scriptTalking pointsSoundings readiness checklistSea Trials sample
Development and governance:
ContributingRelease process (maintainer runbook)Drydock skillsPyPI name reservation notesLaunch distribution plan
Install from a source checkout:
git clone https://github.com/webcloudstudio/Drydock.git
cd Drydock
uv venv
uv pip install -e ".[dev]"
Run local verification:
python -m pytest
ruff check src/ tests/
ruff format --check src/ tests/
Build release artifacts:
python -m hatchling build
After the editable install, run the installed console entry point against the source tree:
drydock --help
Drydock assembles prompts deterministically and runs the selected provider CLI as a subprocess. Execution evidence is persisted under Drydock logs so a run can be audited.
Provider handling is intentionally subscription-oriented:
- API-key environment variables are not the intended execution path.
- Claude and Codex are run through isolated command wrappers.
- Build commands operate against the configured Target workspace and generated application directory.
- Tests use injected runners and never require network or paid API access.
See the "Drydock Security" section in docs/Drydock_Specification.md for the current provider execution contracts.
MIT - Copyright (c) 2026 Web Cloud Studio. See LICENSE.
"Drydock" is a trademark of Web Cloud Studio; see NOTICE for use of the name in forks and derivative works. See CONTRIBUTORS.md for the project's contributors.