Skills & commands
Namespacing
Like the subagents, plugin-provided commands can be
namespaced — /addit-harness:setup is the one that always needs the prefix
(there’s a bare setup collision risk). If a slash command doesn’t respond,
try the addit-harness: prefix before assuming it’s missing.
Skills this repo ships
| Command | What it does |
|---|---|
/adr |
Record an Architecture Decision Record (MADR 4.0) |
/save-plan |
Persist an implementation plan to docs/work/<slug>/plans/ (or --temp) so mermaid renders in an IDE/GitHub |
/go-conventions [--refresh] |
Scan a Go repo and write .claude/go-conventions.md — the project-specific layer on top of the global Go baseline |
/design-conventions [--refresh] |
Scan a TS/React project’s existing UI layer and write .claude/design-conventions.md (tokens, type/spacing/color scales, component lib, layout rhythm, state patterns) |
/addit-harness:setup [--scope global\|project] [--link] [--plugins] |
Places CLAUDE.md/AGENTS.md/rules//references//settings.json for Claude Code — the parts the plugin can’t carry natively. Merges into your existing settings.json (backup first); --plugins also installs the declared plugins. Details in Getting started |
/dev-flow [what to build or fix] [--tier light\|standard\|deep] |
Deterministic SDLC orchestration: triage → tier gate → intake brief → design ⇄ architect-reviewer loop → your approval gate → implement → review ⇄ fix loop → qa-engineer verifies once. --tier overrides the triaged tier. Progress shows in /workflows and, with task tools on, a six-step task list. See How dev-flow works |
/addit-harness:tips |
User-invoked only. Prints five short tips: when to use /dev-flow, approving the plan, where documents land, calling the specialist agents, and the local telemetry option. The first-run welcome line points here. See Getting started |
/addit-harness:telemetry-export [--days N] |
User-invoked only. Builds data.json and an offline report.html from the local telemetry log (default window 30 days); publishes aggregated numbers as a private Artifact only if you answer yes. See Local telemetry |
docs/work/<slug>/solutions/ (architecture designs) and
docs/work/<slug>/architecture-reports/ (review reports) are written directly
by the relevant subagent, not through a
skill — only implementation plans go through /save-plan.
Claude Code’s own built-ins (reused, not reinvented)
The philosophy here is to reuse what Claude Code already ships rather than hand-roll a competing skill:
/code-review— review a diff or PR (see Use cases)./simplify— reduce complexity in existing code./verify— run the project’s tests/build/lint and report real output./run— execute a command with output surfaced back into context./init— bootstrap a newCLAUDE.mdfor a repo.deep-research— multi-step research task.
Official plugins (declared in settings.json)
Enabled from the auto-available claude-plugins-official marketplace (+
anthropics/skills). The big win is real language servers: gopls-lsp,
jdtls-lsp, typescript-lsp, plus pr-review-toolkit, commit-commands,
security-guidance, and document-skills (doc generation). Run
/addit-harness:setup --plugins to install them now.
Iterating & giving feedback on plans or code
The “no way to say change this” problem is mostly a terminal UI gap:
| Scenario | What to do |
|---|---|
| Revising a plan | Choose “Keep planning with feedback” when Claude presents a plan. Type your correction and Claude stays in plan mode. Ctrl+G opens the plan file in your editor to edit directly. |
| Inline comments on code (web) | Open the diff view → click any line → leave a comment. Comments queue and bundle with your next message. |
| Feedback in the terminal | No line-selection UI — reference code as pkg/sales/service.go:45, or paste the relevant lines. Esc interrupts Claude mid-run to redirect it. |
| Browser plan review | Run /ultraplan <task> → open the browser link → leave inline comments → iterate before executing. |