workshop

Tour of the Crucible workshop — the headline orchestrators users invoke directly. Use when the user asks "what skills are available?", "where do I start?", "what should I use for X?", "give me a tour of Crucible", "what are the main commands?", or any onboarding-style question. Also use after onboarding a new user or when someone needs to pick the right tool for a specific task.

10 stars

byraddue

View on GitHub Installation ↓

Best use case

workshop is best used when you need a repeatable AI agent workflow instead of a one-off prompt.

Teams using workshop should expect a more consistent output, faster repeated execution, less prompt rewriting.

When to use this skill

You want a reusable workflow that can be run more than once with consistent structure.

When not to use this skill

You only need a quick one-off answer and do not need a reusable workflow.
You cannot install or maintain the underlying files, dependencies, or repository context.

Installation

Claude Code / Cursor / Codex

$curl -o ~/.claude/skills/workshop/SKILL.md --create-dirs "https://raw.githubusercontent.com/raddue/crucible/main/skills/workshop/SKILL.md"

Manual Installation

Download SKILL.md from GitHub
Place it in .claude/skills/workshop/SKILL.md inside your project
Restart your AI agent — it will auto-discover the skill

How workshop Compares

Feature / Agent	workshop	Standard Approach
Platform Support	Not specified	Limited / Varies
Context Awareness	High	Baseline
Installation Complexity	Unknown	N/A

Frequently Asked Questions

What does this skill do?

Where can I find the source code?

You can find the source code on GitHub using the link provided at the top of the page.

SKILL.md Source

# Workshop

A curated tour of the headline Crucible skills — the orchestrators a user typically invokes directly. Crucible ships ~42 skills total, but most of those are sub-skills that orchestrators dispatch internally. This skill is the **front door**: when you don't know which command to type, start here.

**Skill type:** Reference. Read this skill when the user is choosing what to do; do not dispatch subagents from it.

## Pick by what you're doing

### Starting fresh on something new

| Skill | When to reach for it |
|---|---|
| `/build` | Going from idea to working code. The full pipeline — design → plan → TDD execute → quality gate → PR. One command, complete arc. |
| `/design` | You want to explore design space before any code. Produces a design doc. Build runs this internally; invoke directly when the design itself is the deliverable. |
| `/planning` | You already have a spec or requirements and need a multi-step implementation plan. Build runs this too; invoke directly for plan-only work. |
| `/spec` | You have a GitHub epic with child tickets and want autonomous spec generation (design + plan + contracts per ticket) without human interaction. |

### Mid-feature work

| Skill | When to reach for it |
|---|---|
| `/temper` | Iterative code review on a PR or `<base>..<head>` range. Loops fresh-eyes reviewers until clean. Forge-agnostic (GitHub / GitLab / Bitbucket / self-hosted). Renamed from `/code-review` on 2026-05-17. |
| `/debugging` | Any bug, test failure, or unexpected behavior. Hypothesis loop, fix dispatch, verification. |
| `/quality-gate` | Iterative red-team on any artifact (design doc, plan, code, hypothesis, mockup). Loops until clean or stagnation. Invoked by other orchestrators; invoke directly when the artifact is standalone. |
| `/migrate` | Framework upgrade, API version bump, major dependency change, deprecation removal. Produces a phased migration plan and optionally executes it via build's refactor mode. |

### Inspection and discovery

| Skill | When to reach for it |
|---|---|
| `/delve` | Instance-bug review of a diff or path — parallel finder angles + verify gate, prints ranked, verified defects with reproductions. Report-only (no merge verdict, no fix loop; `--fix` / `--comment` are opt-in). Use when "find bugs in this diff", "scan this file for defects", or you want concrete reproducible bugs. |
| `/audit` | **Systemic** review of an existing subsystem or non-code artifact (design / plan / concept) — recurring patterns, structural drift, absences with no single reproduction. Four lenses per artifact type, find-and-report only (does not fix). Instance bugs route to `/delve` via `--bugs`; complexity to `/prospector`. Use when "this has accumulated cruft nobody's looked at in months." |
| `/siege` | Security audit. Six parallel attacker-perspective Opus agents, iterates until zero Critical/High. Heavy — reserve for security PRs, scheduled reviews, or post-incident. |
| `/recon` | Codebase investigation. Layered Investigation Brief with structure / patterns / scope / prior-art. Use before any task that needs codebase understanding you don't have. |
| `/prospector` | Architectural friction finder. Explores the codebase for refactor candidates and proposes competing redesigns. Use when "what should I refactor next?" |

> **`/delve` vs `/recon` / `/prospector` — different machines.** `/delve` hunts concrete *instance bugs* in a specific diff or path (one defect, one reproduction). `/recon` and `/prospector` *explore* unfamiliar code — recon maps structure / patterns / prior-art before a task; prospector finds architectural friction and proposes redesigns. Reach for `/delve` when you have a **change** and want its bugs; reach for recon / prospector when you have a **codebase** and want to understand or improve it. Within the review trio: `/delve` = instance bugs (one-shot), `/audit` = systemic patterns, `/temper` = merge gate + iterative fix loop.

### Wrapping up and reflecting

| Skill | When to reach for it |
|---|---|
| `/finish` | Implementation is complete, tests pass, and you need to decide how to integrate — merge, PR, cleanup. Guides the completion decision. |
| `/forge` | A significant task just completed. Writes a retrospective; proposes skill mutations after 10+ retrospectives accumulate. Compounding knowledge accelerator. |
| `/handoff` | End-of-session. Writes a handoff doc for the next session — continuation of current arc, or proposed next pickup if the current arc is wrapping. |

## Quick reference by trigger phrase

| User says… | Reach for |
|---|---|
| "Let's build X" / "Implement Y" | `/build` |
| "Review my PR" / "Code review" | `/temper` |
| "Find bugs in this diff" / "Scan this file for defects" | `/delve` |
| "I have a bug" / "Test is failing" | `/debugging` |
| "Audit this design" / "Review this plan" | `/audit` |
| "Check the save system for bugs" | `/audit` (systemic) or `/audit --bugs` (+ `/delve` instance sweep) |
| "Security review" / "Threat model" | `/siege` |
| "What does X do?" / "How does this codebase work?" | `/recon` |
| "What should I refactor?" | `/prospector` |
| "Run quality gate on this" | `/quality-gate` |
| "Upgrade React to 19" / "Remove deprecated APIs" | `/migrate` |
| "I'm done, what's next?" | `/finish` |
| "Write a retrospective" | `/forge` |
| "End of session, write a handoff" | `/handoff` |

## Pipelines (composition)

The big skills compose. The most common pipelines:

- **Full feature:** `/build` is the canonical one. Internally runs `/design` → `/planning` → execute → `/temper` per task → `/inquisitor` → `/quality-gate` → `/finish`.
- **Standalone design + plan:** `/design` produces a doc → user reviews → `/planning` produces a plan → execute manually or feed to `/build`.
- **Post-merge reflection:** `/forge` after a significant arc; `/handoff` at session boundary.
- **Recon-first:** `/recon` before `/build` or `/design` when the codebase is unfamiliar.

## What's not in this list (intentionally)

Crucible has ~42 skills; this list curates ~15. Skills omitted here are either:
- **Sub-skills** dispatched by orchestrators (e.g., `red-team`, `inquisitor`, `adversarial-tester`, `checkpoint`, `verify`, `assay`, `innovate`, `cartographer-skill`) — invoked indirectly, not part of the user-facing menu.
- **Domain-specific** (e.g., `mock-to-unity`, `ui-verify`, `mockup-builder`) — load when their domain triggers.
- **Utility / meta** (e.g., `skill-creator`, `stocktake`, `recall`, `replay`, `worktree`, `parallel`, `merge-pr`, `distill`, `test-coverage`, `review-feedback`, `consensus`, `getting-started`, `project-init`, `test-driven-development`) — domain knowledge or workflow plumbing rather than direct-invocation commands.

Run `/skills` (Claude Code built-in) for the full catalog.

## See also

- `README.md` — top-level pitch, install, what-you-get bullets
- `docs/architecture.md` — how the orchestrators compose
- `docs/skills.md` — full 42-skill catalog with eval deltas

Related Skills

worktree

from raddue/crucible

Use when starting feature work that needs isolation from current workspace or before executing implementation plans - creates isolated git worktrees with smart directory selection and safety verification

verify

from raddue/crucible

Use when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always

ui-verify

from raddue/crucible

Use when verifying Unity UI matches a mockup, after implementing UI from a visual reference, when a user shares a screenshot showing UI drift, when checking theming compliance, or when UI looks wrong. Triggers on "verify", "compare to mock", "does this match", "check the UI", "screenshot shows wrong", "UI looks wrong", "fix the layout", "spacing is off", "colors are off", "doesn't look like the design", "visual bug", or completing any UI implementation task.

test-driven-development

from raddue/crucible

Use when implementing any feature or bugfix, before writing implementation code

test-coverage

from raddue/crucible

Audit existing tests for staleness, needed updates, or removal after code changes. Use after any code modification to verify test suite alignment. Triggers on 'audit tests', 'check test alignment', 'stale tests', 'test health', or any task verifying tests match changed code. Technology-agnostic — works with any language and test framework.

temper

from raddue/crucible

Iteratively review code changes for production readiness through fresh-eyes review loops. Use when completing tasks, implementing major features, or before merging — including when the user says "review this PR", "review my changes", "code review", "check the diff", or "is this ready to ship". Works on PRs from any forge (GitHub, GitLab, Bitbucket, self-hosted) or on raw git SHA ranges.

temper-eval-collect

from raddue/crucible

Live-dispatch phase of temper eval harness. Reads stage-manifest.json from a pre-staged dispatch dir; fans Task-tool reviewer dispatches in parallel (max 6); writes per-seq result files; exits. Single bounded session. Pairs with `python -m skills.temper.evals.run_evals stage` and `score`.

temper-eval-calibrate

from raddue/crucible

Wrapper skill that runs k iterations of (stage → collect-behavior → score) for

stocktake

from raddue/crucible

Audits all crucible skills for overlap, staleness, broken references, and quality. Quick scan or full evaluation modes.

spec

from raddue/crucible

Use when you have a GitHub epic (or equivalent) with child tickets and want to autonomously produce design docs, implementation plans, and machine-readable contracts for each ticket without human interaction. Triggers on /spec, 'spec out', 'write specs for', 'spec this epic'.

source-driven-development

from raddue/crucible

Enforces the Detect → Fetch → Implement → Cite protocol when implementing against external frameworks or libraries. Invoke when a change touches an external API surface and the edit exceeds the triviality threshold, so implementations come from current official docs rather than stale training-data recall.

skill-selection-evals

from raddue/crucible

Eval-only skill for measuring skill routing accuracy. Not invoked directly — contains selection evals that test whether the agent picks the correct skill for a given prompt.