User story
An AI agent initializes a dotdog project or audits issue coverage without
running CLI commands. Two new MCP tools expose the same features.
New tools
initGuided — non-interactive project init from parameters
{ name: 'initGuided',
description: 'Initialize a dotdog project from parameters',
inputSchema: { properties: {
project: { type:'string' },
purpose: { type:'string' },
users: { type:'string' },
rules: { type:'string' },
entities: { type:'string' }
}, required: ['project'] } }
checkIssues — validate spec coverage against GitHub issues
{ name: 'checkIssues',
description: 'Validate spec coverage against GitHub issues',
inputSchema: { properties: {
repo: { type:'string' },
project: { type:'string' }
}, required: ['repo'] } }
Implementation
- Add tool defs to
tools/list handler in serve.ts
- Add handlers in
tools/call — reuse logic from CLI
- Extract shared
initProject() and checkIssues() from cli.ts
Scope
- 2 new MCP tool definitions
- 2 new if-blocks in tools/call handler
- ~30 lines in serve.ts, ~10 line refactor in cli.ts
- Zero new files, zero new deps
Depends on: init --guided, issues --check
User story
An AI agent initializes a dotdog project or audits issue coverage without
running CLI commands. Two new MCP tools expose the same features.
New tools
initGuided— non-interactive project init from parameterscheckIssues— validate spec coverage against GitHub issuesImplementation
tools/listhandler in serve.tstools/call— reuse logic from CLIinitProject()andcheckIssues()from cli.tsScope
Depends on: init --guided, issues --check