thomaslty/claude-skills

Custom slash commands and plugins for Claude Code.
## Components
### Skills
Skills live in skills/. Copy to install:
bash
cp -r skills/explore ~/.claude/skills/explore
cp -r skills/no-code-change ~/.claude/skills/no-code-change
cp -r skills/propose ~/.claude/skills/propose
cp -r skills/review ~/.claude/skills/review
| Skill | Description |
| /explore | Read-only thinking partner — explore ideas, plan approaches, and stress-test designs without touching code |
| /no-code-change | Show-first workflow — gather, summarize, and present before making any changes |
| /propose | Sequential 3-agent pipeline (explore, engineer, verify) — vetted solution proposals with online fact-checking, read-only |
| /review | Comprehensive code review with 6 parallel sub-agents (quality, reuse, truthfulness, efficiency, security, best practices) |
### Commands
Slash commands live in commands/. Copy to install:
bash
cp commands/commit.md ~/.claude/commands/commit.md
cp commands/show-plan.md ~/.claude/commands/show-plan.md
cp commands/fable-design-opus-execute.md ~/.claude/commands/fable-design-opus-execute.md
cp -r commands/dev ~/.claude/commands/dev
cp -r commands/hfv ~/.claude/commands/hfv
cp agents/.md ~/.claude/agents/
mkdir -p ~/.claude/templates && cp -r templates/dev ~/.claude/templates/dev
mkdir -p ~/.claude/scripts && cp -r scripts/dev ~/.claude/scripts/dev
The dev:0-explore … dev:full-dev-cycle commands need the agents/ (one fresh agent per phase) and templates/dev/ (the decision page) copies too.
| Command | Description |
| /commit | Git commits without the Co-Authored-By trailer |
| /show-plan | Show the full current plan — goal, dependency graph, matrix table, worked detail per step |
| /fable-design-opus-execute | Fable designs in the main session; Opus sub-agents execute the code |
#### dev: — development lifecycle and code review
| Command | Description |
| /dev:help | Quick-reference card for the dev: namespace |
| /dev:development-guideline | Mandatory methodology — proposal-driven, TDD backend, visual-driven frontend |
| /dev:parallel-development | Build multiple proposals at once via per-worktree parallel sub-agents |
| /dev:finishing-development | Archive, then land each repo by its own .claude/finishing-development.md; asks for the rules first when a repo has none |
| /dev:ticket | Orca board ticket — new makes a card + worktree, start launches Claude on it |
| /dev:0-explore | Fresh agent expands and grills a rough idea round by round (research, Claude Design canvas, HTML page) until it settles into plan/ /ticket.md |
| /dev:1-design | Fresh agent writes plan/ /plan.md, asks you via a decision-tree HTML page |
| /dev:2-implement | Fresh agent follows the plan under the guideline until every step is verified |
| /dev:3-review | 4 fresh reviewers (correctness, security, feature, completeness), T0 only, HTML page |
| /dev:full-dev-cycle | Design ⏸ → implement → review ⏸ → fix, each phase a fresh agent, stops for you after design and review |
| /dev:learn-conventions | Merge learned coding conventions into .claude/conventions.md |