A CLI tool for managing AI-assisted development workflows, including work state management, memory persistence, legacy .wip migration, and repository bootstrapping
This server has 23 tools with significant definition gaps. Most tools lack proper input schemas (cannot verify JSON Schema structure in source), descriptions are present but often generic or under-specified, and parameter types are not consistently documented. Tools like 'bootstrap', 'cleanup-branch', and 'migrate-state' have destructive/write side effects but lack confirmation patterns or dry-run guidance in descriptions. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) are visible. Error handling is not documented. Output schemas are not specified. The server reads as a CLI wrapper rather than an LLM-optimized tool interface.
Initialize the global aidw environment and optionally bootstrap a repository
Bootstrap all repos in a workspace
Archive all files in the current branch .wip dir except context.md, pr.md, spec.md, task-context.md and status.json
Archive all .wip branch dirs except the current branch's dir
Archive all .wip branch dirs except the most recently dated one
Print context-summary.md to stdout, or its staleness as JSON with --json
Generate/Update project documentation in .claude/repo-docs/
Bootstrap a single repo
No input schemas visible in source code. Tools declare parameters (path, dry-run, etc.) but JSON Schema structure with type definitions, constraints, and enums is not evident. Cannot verify schema completeness.
Destructive operations (cleanup-branch, clear-wip, clear-others, migrate-state with --cleanup-sources) lack confirmation patterns or explicit dry-run guidance in descriptions. LLMs cannot safely invoke these without risk of data loss.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) detected. LLMs cannot determine which tools are safe to retry, which modify state, or which are read-only without explicit hints.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 37 | 2026-07-28+ | v2 |
Generate .github/agents/ from claude/agents/ stripping MCP sections
Generate .github/skills/ from claude/skills/ (byte-identical mirror)
PreToolUse hook: gate Edit/Write/NotebookEdit behind an active .wip session
Index documentation for semantic search
List all persistent facts. Defaults to current branch.
Rebuild the memory database onto the repo_id/scope schema
Perform semantic search over project knowledge
Check the health of the memory layer
Store a persistent fact for the current branch
Insert or replace the managed block in CLAUDE.md
Add or update Serena and Context7 in ~/.claude/mcp.json
Deep-merge a JSON template into settings.json
Migrate legacy .wip branch state into the work-model store
Generate context-summary.md from all WIP files
Add managed lines to the global gitignore
Output schemas not documented. Tools like 'context-summary', 'memory search', and 'memory list' return data but the response structure is not specified, forcing LLMs to guess field names and types.
Error handling and recovery guidance absent. No tool description explains what to do if a call fails, whether errors are retryable, or what the LLM should try next.