From groundwork
One-time per-project setup — installs git hooks, commit template, WORKING_AGREEMENT.md, and CLAUDE.md
npx claudepluginhub vohaha/groundwork --plugin groundworkThis skill uses the workspace's default tool permissions.
Install groundwork git integration for this project.
Designs and optimizes AI agent action spaces, tool definitions, observation formats, error recovery, and context for higher task completion rates.
Compares coding agents like Claude Code and Aider on custom YAML-defined codebase tasks using git worktrees, measuring pass rate, cost, time, and consistency.
Designs, implements, and audits WCAG 2.2 AA accessible UIs for Web (ARIA/HTML5), iOS (SwiftUI traits), and Android (Compose semantics). Audits code for compliance gaps.
Install groundwork git integration for this project.
The plugin root is: ${CLAUDE_PLUGIN_ROOT}
Steps:
Run the setup script: bash "${CLAUDE_PLUGIN_ROOT}/scripts/setup.sh" This installs git hooks and commit template deterministically.
Create or check WORKING_AGREEMENT.md at the project root:
[project] in the header with the actual project nameCreate or update CLAUDE.md:
<!-- groundwork:start --> marker
or ## Groundwork heading): replace it with the template contentThe groundwork section is owned by groundwork — replace it fully on re-run, do not merge.
Seed user-owned sections in CLAUDE.md (skip each if already exists):
a. Operating Mode (skip if ## Operating Mode exists):
Prepend before the groundwork section.
## Operating Mode
Claude drives. The human's role is to support — clarifying intent, reviewing decisions,
and providing context Claude doesn't have. Not the other way around.
Operational implications:
- Claude acts on obvious improvements without waiting for permission
- Honest pushback is preferred over polite compliance
- When evaluating a change, the first question is: does this reduce friction or move the work forward?
b. Domain (skip if ## Domain exists):
Add after Operating Mode, before the groundwork section.
## Domain
<!-- Brief domain summary — what a new session needs to orient.
For detailed domain knowledge, see docs/domain/.
This section is user-owned — groundwork will not overwrite it.
Update during check-in when domain understanding deepens. -->
Create docs/domain/ directory if it doesn't exist:
mkdir -p docs/domain
If the directory was just created (empty), add a README.md:
# Domain Knowledge
Detailed domain context for Claude. Files here are read by the plan skill
before designing features — they provide the expertise a new session needs.
Keep files focused: one concept per file, written for an agent that has
never seen this domain before.
Examples:
- `game-mechanics.md` — how the core game systems work
- `user-workflows.md` — what users actually do with this app
- `terminology.md` — domain-specific terms and their meanings
Run validation: bash "${CLAUDE_PLUGIN_ROOT}/scripts/validate.sh" If any checks fail, fix them before reporting.
Report what was installed, any steps skipped, and validation results.
$ARGUMENTS