Repository navigation
Releases: iowarp/wtf-ms
Release list
wtf-MS 1.0.0
First tagged release. wtf-MS is now a Claude Code plugin, and the same workflow ships as a Clio Coder extension.
Install
claude plugin marketplace add iowarp/wtf-ms
claude plugin install wtfMS@wtf-msRestart Claude Code, then run /wtfMS:help. At user scope the 18 /wtfMS:* commands are available in every project.
Breaking changes
The copy-into-your-project install is gone. Previously you copied three directories into YOUR_PROJECT/.claude/. Templates, references, and guardrail scripts are now addressed through ${CLAUDE_PLUGIN_ROOT}, so they resolve against the installed plugin instead of your working directory. Use the plugin install above.
Metrics hooks are no longer installed by hand. Stop and SubagentStop ship with the plugin in hooks/hooks.json. The previous setup told you to add them to your project's .claude/settings.json pointing at $CLAUDE_PROJECT_DIR/.claude/wtf-ms/scripts/record_metrics.py, a path that no longer exists once the script lives inside the plugin. record_metrics.py returns early unless the working directory contains a .research/, so it is inert outside a wtf-MS project. If you added those hooks manually, remove them to avoid double-counting.
Layout
The repository root is the plugin root. commands/ and agents/ moved up from .claude/commands/wtfMS/ and .claude/agents/wtfMS/; resources moved from .claude/wtf-ms/ to wtf-ms/. Agent files are renamed to wtfms-<slug>.md — Claude Code derives a plugin agent's identity from its filename rather than its frontmatter name, and the linter now enforces the match.
Clio Coder
The same workflow ships as a hand-written extension under clio/: the 18 prompts, six agent recipes, bound skills, an execute-and-verify fleet, and the references, templates, and guardrail scripts as extension resources.
clio-coder extensions install ./clio --userAlso in this release
- MIT
LICENSEfile, matching the license the README has always declared. - Guardrails carried forward: Crossref citation verification, physical-sanity checks, and runnable-by-construction script checks, all advisory.
- Token/cost metrics and the
eval/benchmark harness.
Notes
The plugin is named wtfMS rather than kebab-case so the command namespace stays /wtfMS:*. claude plugin validate warns about this; it only affects Claude.ai marketplace sync, not local installs.