mold — Claude Code skill: what it does, how to install
mold turns a high-stakes design choice into a checked specification through guided dialogue. Use it before architectural or destructive changes.
npx skills add paulnsorensen/easy-cheese --skill mold -a claude-codeWhat mold does
The skill frames the goal, separates settled points from open forks, and selects a dialogue mode such as Explore, Ground, Shape, Sketch, Grill, or Diagnose. It checks claims against code and project evidence, defines module boundaries, and records the decisions that shape the specification.
Its procedure has Quick, Light, and Full tiers, reference material under references/, coherence checks, and strict spec validation. The durable artifact path comes from python3 skills/mold/scripts/mold.pyz artifact-path specs <slug>. Once the draft is reconciled, mold can offer a handoff to Cook.
The taste-test check requires python3; without it, python3 skills/mold/scripts/mold.pyz taste-test cannot run. /culture, /cook, and /briesearch cover free-form discussion, direct implementation, and research-only questions respectively.
What it helps with
- turn a risky design choice into a written spec
- check the impact of authentication or irreversible operations
- settle a new public interface
- map dependencies across modules
- save a validated spec for Cook
Example prompts for the agent
- Run
/moldand help me design authentication for this project. - Shape the API for a new public interface, then test the unresolved design forks.
- Use
/moldto work through an irreversible operation and save the resulting specification.
How to install mold
1Skills CLI (Vercel)
npx skills add paulnsorensen/easy-cheese --skill mold -a claude-code2Manually: put the files in place
~/.claude/skills/mold/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/mold/
1Skills CLI (Vercel)
npx skills add paulnsorensen/easy-cheese --skill mold -a codex2Manually: put the files in place
.agents/skills/mold/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add paulnsorensen/easy-cheese --skill mold -a cursor2Manually: put the files in place
.cursor/skills/mold/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install paulnsorensen/easy-cheese mold2Manually: put the files in place
.github/skills/mold/SKILL.md1Skills CLI (Vercel)
npx skills add paulnsorensen/easy-cheese --skill mold -a gemini-cli2Platform CLI
gemini skills install https://github.com/paulnsorensen/easy-cheese --path skills/moldManually: put the files in place
.devin/skills/mold/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add paulnsorensen/easy-cheese --skill mold -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about mold
How do I install mold in Claude Code?
Run npx skills add paulnsorensen/easy-cheese --skill mold -a claude-code, then check ~/.claude/skills/mold/SKILL.md.
What does mold do?
It guides design decisions through grounded dialogue, checks the reasoning behind them, and saves a validated specification. The result can then be handed to Cook.
How is mold different from /cook and /culture?
mold is for shaping a design into a specification. /cook handles direct implementation, while /culture is for discussion without an artifact goal.
Source text
SKILL.md · MIT
---
name: mold
description: Turn a high-leverage design decision into a validated spec through grounded dialogue. Use it when a leverage trigger fires (auth, irreversible effects, concurrency, contracts, destructive ops, a new slice, a cross-slice dependency, an invariant gap) or when the user asks for a spec or design. Typical phrases include "let's design X", "shape this into a spec", "what should the API for Z look like", and "/mold". A feature ask with zero fired triggers is a `/cook` mini-spec, not a mold; route it through `/cheese`. Do NOT use it for free-form discussion without artifact intent (`/culture`), direct implementation (`/cook`), or research-only questions (`/briesearch`). Ceremony scales to the ask through `references/tiers.md`, so a small clear change gets a validated mini-spec without a write-approval turn.
license: MIT
metadata: {dispatches-agents: true}
---
# /mold
Ceremony scales to the job. The Bounds pass picks one of three tiers from `references/tiers.md`: **Quick** writes a validated mini-spec, **Light** runs only the dialogue modes the open forks need, and **Full** runs the whole Flow below. Saving a spec needs no extraction approval. Cook still needs an explicit user request. `/cheese`'s tier-1 escalation enters mini-spec mode; see `## Agent-invoked mini-spec mode`.
## Flow
1. **Bounds pass** — open the `Goal / Decided / Asking / [AGENT-DECIDED]` ledger with the goal pinned as one sentence; only an explicit user fork changes it. Split it into `G-n` clauses (`references/handshake.md` § Goal coverage). Map every input's goals and **non-goals** as one `[AGENT-DECIDED]` line; ask the user only when the goal is genuinely unknown or a leverage trigger fires. Run the shape check, then announce the tier with its reason (`references/tiers.md`). Quick exits here to `## Agent-invoked mini-spec mode` without a write-approval turn. Upgrade the tier whenever the evidence changes; never downgrade silently.
2. **Route** — choose the secondary mode from `references/modes.md`, announce it, and correct false premises first.
3. **Dialogue** — consequential forks are the user's to pick. A fork is consequential per the leverage line in `../age/references/voice.md`. Every other fork is `[AGENT-DECIDED]`. Supply options, trade-offs, and evidence before you ask. Ground each critical claim through code, the [Validate Cycle](references/validate-cycle.md), or a [Prototype Cycle](references/prototype-cycle.md). Resolve every contradiction. Render the decision map after three consecutive fork questions, or on request.
4. **Sketch** — For work across modules or with a new public interface, run `references/shape-check.md`. Bind identity and role nouns to code referents. Record the Placement block; no bodies.
5. **Plan and validate** — run the fresh-context fork-coherence taste test with `python3 skills/mold/scripts/mold.pyz taste-test`; a failure reopens only named forks, and the third failed verdict stops. The taste review guides the spec but is not a Cook execution reference. For bound curd work, dispatch a typed `PlannerRequest`, validate its `PlannerResultWriterView` (one retry, then save without runnable inputs), and normalize on the host. A clear single-curd task can instead use standalone Cook without a planner. Show semantic curds and waves before any Cook choice. See `references/curdle.md` § "Typed planner dispatch".
6. **Readiness check** — Before saving, run the coherence self-check and strict spec validation. Save a draft when unresolved parent work remains; record each hold in `execution_holds:`. Finalization refuses ready while an entry remains. User approval is not a prerequisite for writing. Before Cook, require an explicit user selection of scope and route. See `references/handshake.md` and `references/early-curds.md`.
7. **Curdle** — Resolve the durable spec path with `python3 skills/mold/scripts/mold.pyz artifact-path specs <slug>` and read `path` from its JSON reply. Write the validated draft and read it back. Retain typed plan evidence and unresolved holds. Write local issue drafts and durable decisions; publish follow-ups only after their own approval. A concrete bounded curd may use `references/early-curds.md` while Mold continues shaping its parent. The selected spec path can enter standalone Cook; bound curd execution also receives validated direct refs, never a pointer.
8. **Offer Cook or keep shaping** — after reconciliation, run [`python3 skills/mold/scripts/mold.pyz curd-count`](references/curd-count.md). A saved draft is not an execution request. If the user selects bound curd execution, run `scripts/mold.pyz finalize` and pass only ready direct inputs. A standalone clear task can enter Cook by spec path without machine finalization. Otherwise keep shaping or stop without a Cook command.
Portability: [rules](../cheese/references/harness-portability.md). Slash commands are host renderings, not the control model.
## Follow-up candidates
Every non-goal and explicit dialogue deferral becomes a `[FOLLOW-UP?]` follow-up candidate. Dispose of the set before a runnable Cook handoff; details: `references/handshake.md` § Follow-up disposition.
## Modes
| Mode | Use when | Goal |
| --- | --- | --- |
| Explore | The idea is vague | Identify the real problem and pain point |
| Ground | A file, bug, or existing doc is named | Verify facts against evidence |
| Shape | The goal is known but approach is open | Compare viable options (Do Nothing always included) |
| Sketch | Interfaces or module boundaries matter | Lock responsibilities and seams |
| Grill | A favoured approach needs stress-testing | Steelman each item, then put every design-changing call to the user as a fork |
| Diagnose | A symptom, failure, or trace is supplied | Build a Loop → reproduce → hypothesize → confirm root cause |
Modes: references/modes.md. Evals: references/evals.md. Canvas: [review-canvas.md](references/review-canvas.md).
## Agent-invoked mini-spec mode
`/cheese`'s tier-1 escalation calls `/mold` after the call site passes all Cook fast-path checks. The Quick tier enters the same mode without a write-approval turn. It produces a validated draft without design dialogue. Return the spec path. After the spec validates, show it and ask one direct Cook question. On explicit selection, dispatch `/cook --spec <path>` for clear standalone work; do not invent a pointer or require a planner. Append `--hard` when the user passed it.
The full coherence checklist does not run in this mode. The agent-introduced-scope check still runs implicitly. Every distinguishing noun must come from the user's input or tier-2 `/culture`/`/briesearch` synthesis. Never add one silently.
Full procedure, the mini-spec schema, and the `## Provenance` rules: `references/mini-spec-mode.md`.
## Preferred tools and fallbacks
Call source-code search, read, and edit backends according to [`code-intelligence-routing.md`](../cheese/references/code-intelligence-routing.md). Shape checks use semantic caller search plus dependency context; procedure: `references/shape-check.md`.
Mold-specific tools beyond source-code routing:
| Need | Prefer | Fallback |
| --- | --- | --- |
| External validation | `/briesearch` with Context7/Tavily | user docs, repo docs, or note as unverified |
| Wiki grounding (Ground entry + decision points; scope per `references/grounding.md` § When to probe) | `mcp__hallouminate__list_corpora` + `mcp__hallouminate__ground` on `repo:<repo>:wiki` | record `hallouminate: absent` in the ledger, proceed on code evidence, and cap at `speculating` when design rationale is central |
**The grounding record precedes the first structured question:** the ledger holds a probe result — citations or `hallouminate: absent` — first. Mark each unsupported claim `[?]` until settled.
## Sub-agent context gate
`/mold` owns the dialogue, contradictions, and save-approval state. Do not delegate these items. Delegate evidence-heavy code work to a fresh-context `explorer` and external research to a `researcher`. **Shape uses an explorer digest as input.** Record parent-context exploration as a degraded path. See `references/context-budget.md` for budgets and checkpoints.
### Gate graph
`python3 skills/mold/scripts/mold.pyz gate-graph --render dot|svg|png|mermaid` renders one gate model. Image targets use Mermaid when Graphviz is unavailable. Tests keep gate nodes aligned with the handshake checklist. See `references/gate-graph.md`.
### Spec provenance and Test Contracts
Every Mold-produced spec carries a provenance marker in frontmatter:
```yaml
source: mold-handshake | agent-mini-spec | mold-curd-mini-spec
```
A spec has no `disposition` field. `## Test Contracts` is optional: a spec may
include it, and no gate requires it. Row-level rules: `references/curdle.md`
§ Test Contracts.
### Fork taste gate
`python3 skills/mold/scripts/mold.pyz taste-test` binds the verdict to draft SHA256 and each settled fork. Stale, partial, or blocked verdicts fail; a failure reopens only named forks, with two rounds. Approved specs pass unchanged metadata to `/cook --spec <spec-path>` (add `--auto` only when the user chose it) for standalone work; bound curd work uses `/cook --auto --result <result-path>`.
Each fork appears in Approach, Interface sketches, and Acceptance, plus Test Contracts when the draft has them. Do not rename sections. Tag reflecting lines with fork id and run `taste-test --precheck` before dispatch. `goal` must survive verbatim in Problem statement (`goal-drift`); each `G-n` clause carries an Acceptance or disposition tag (`goal-coverage`). See `references/curdle.md` § Spec template and `references/gate-graph.md`.
## Execution gate
Mold may save a validated parent spec or early curd mini-spec without an approval turn. The agent coherence check still controls what the artifact claims as settled. Print the narrowing delta (`taste-test --coverage`) and the scope audit table before presenting any Cook option. Unresolved decisions remain visible holds, not implicit approvals. See `references/handshake.md`.
Never start Cook from a saved spec, silence, `--auto`, or a general design approval. The user must select the exact curd or parent scope and a Cook route. A changed scope or plan needs a new selection. A failed bound finalization leaves a saved draft without executable refs. See `references/early-curds.md`.
## --hard
Mold never runs the metacognitive check. Plate alone runs it at the
verified-artifacts boundary. Mold appends `--hard` to every Cook command that
it emits when the user passed the flag. Cure forwards the same flag to Plate.
See `../hard-cheese/references/composition.md`.
## Handoff
**Pipeline:** culture → **[mold]** → cook (⇄ press) → age → cure → plate
After Curdle's phase two finishes, run [`python3 skills/mold/scripts/mold.pyz curd-count`](references/curd-count.md) for the blast-radius digest. A draft offers no Cook command. On explicit Cook selection, retain the selected spec path for standalone work. Bound curd work dispatches the printed finalize command, `/cook --auto --result <result-path>`; the result carries `request_id`, `spec_ref`, `planner_result_ref`, `plan_ref`, and exact `coverage`. Prompt through the shared handoff gate ([policy](../cheese/references/handoff-gate.md)); never pre-select. Keep applicability, contract, and taste review metadata unchanged. Append `--hard` when the user passed it.
The digest's `mode` is orientation, not a skill or Cook gate. Render the fixed blast-radius menu from `decomposable`, `candidate_curds`, `verdict`, and `mode`; see `references/handoff-menus.md`.
## Rules
- Dialogue first; artifacts are the by-product.
- **Tiered lettered options.** Consequential forks use `A/B/C/D` choices via the question transport at `../cheese/references/ask-user-question.md`. Never decide them silently. Everything below the leverage line is `[AGENT-DECIDED]` by default: make the call, log a one-line vetoable alternative in the ledger, and do not ask. A fork is valid only after its depth was contributed in-dialogue first. Precede every structured question with visible prose that weighs the fork and the evidence. Keep one open picker.
- **Fork cards.** Before each consequential question, show a fork card: what it decides, what it moves, and each branch as a symbol table or plain-text graph. Offer `Explain more` as a choice; an unclear reply re-renders the same fork with more context ([fork-questions](references/fork-questions.md)).
- **Altitude tag.** Every `Asking` fork names the acceptance criterion, public seam, or non-goal it moves. A fork that moves none, or that sits below the leverage line, is `[AGENT-DECIDED]` or a follow-up candidate, never a user question.
- **Decision ledger.** Each round prints `Goal / Decided / Asking / [AGENT-DECIDED]`, the goal verbatim. Curdle persists consequential decisions to [ADRs](references/adr.md) and minor ones to the spec. The taste verdict names every settled consequential entry exactly once.
- **Decision map and fork-round cap.** Stop after three consecutive fork rounds that add no new evidence, or three consecutive forks that fail the altitude tag, or when the user requests a decision map. Show completed forks, remaining forks, and a ready or blocked verdict. A fourth round requires new grounding, a delegated digest, or a `/wheypoint` checkpoint. The map shows ledger state. It does not create an artifact.
- Do not implement code.
- Do not write production files in Mold. Saving a spec does not authorize Cook.
- Do not silently settle uncertain claims.
- Apply the shared voice kernel at `../age/references/voice.md`. Correct false premises. Mark each critical claim's confidence as `certain | speculating | don't know`. Steelman before you dismiss. Put consequential forks to the user; depth informs the question, never replaces it.
The generated [`../cheese/references/schema-intertwine.md`](../cheese/references/schema-intertwine.md) lists every published schema and its contract model.
## Agent resolution
Resolve delegates through [`../cheese/references/agent-resolution.md`](../cheese/references/agent-resolution.md).
| Work | Preferred types | Permissions/isolation | Minimum power | Effort | Fallback |
| --- | --- | --- | --- | --- | --- |
| Explore the codebase | explorer | read-only, fresh-context | default | medium | compatible explorer, then general |
| Research external constraints | researcher | read-only, fresh-context | default | medium | compatible researcher, then general |
| Plan for approval | planner, general | read-only, fresh-context | powerful | high | compatible planner, then general |
The canonical mold spec or mini-spec carries the shared `agent_resolution` block.
Commands: [`references/commands.md`](references/commands.md).
© paulnsorensen · MIT · paulnsorensen/easy-cheese