feat(icp-cli): use mops generate candid; Rust recipe v3.3.0 (optional package)

[marc0olo]

Closes #214.

Improves the icp-cli skill on two fronts: adopting mops generate candid for Motoko Candid generation, and updating the Rust recipe to v3.3.0 (where package is now optional).

Changes to skills/icp-cli/SKILL.md

Issue #214 — mops generate candid

• Pitfall 16 (Motoko binding-generation path) now recommends mops generate candid backend — a single step that extracts the Candid interface directly from Motoko source without compiling WASM — replacing the old mops build backend + cp .mops/.build/backend.did … two-step. Cross-references the mops-cli skill, where the command is already fully documented. The "nothing to do" path (no binding generation needed) is unchanged, as the issue requested.

Rust recipe v3.3.0 — package now optional

• Bumped all @dfinity/rust@v3.2.0 → v3.3.0 references (pitfalls 3, 4, 14; Rust config example; recipe table).
• Recipe table: package moved from required → optional (defaults to the canister name), and added the previously-missing metadata optional config.
• New Pitfall 21: package is optional when the canister name matches the Cargo [package] name; set it only when they differ (e.g. a workspace crate). Pitfall 4 and the Rust config section cross-reference it.

Evals (evaluations/icp-cli.json)

Added two cases and re-scoped one. All changed/added cases were run with baseline.

Eval results

Case 24 — Candid generation for Motoko binding (mops generate candid) (new) — WITH 5/6 | WITHOUT 2/6

The one WITH miss is a completeness nuance (the "overwrites the existing [canisters.backend].candid path if set" branch); the prompt described a fresh setup so the model covered the other branch. Both branches are documented in the skill body.

Case 25 — Adversarial: Rust recipe package parameter (minimal vs workspace) (new) — WITH 4/4 | WITHOUT 4/4

No with/without delta: Opus 4.8 already knows the v3.3.0 package convention and reasons it out. Kept as a regression guard that pins the exact version behavior (and would catch a future WITH-skill regression / weaker models).

Case 15 — Full-stack Motoko config artifacts (reworked) — WITH 6/6 | WITHOUT 4/6

The previous version of this case asked for an entire full-stack project in one prompt (icp.yaml + mops.toml + Motoko source + bindings + Vite dev-server + React component + walkthrough). That response exceeded the eval runner's hard 120s subprocess timeout, so claude was killed with ETIMEDOUT and every behavior failed for both WITH and WITHOUT. Re-scoped the prompt to just the config artifacts + the candid command (excluding React component code and prose), and dropped the dev-server/outDir/local-network behaviors that are already covered by cases 19 and 6. It now completes reliably and shows the mops generate candid delta.

🤖 Generated with Claude Code

[github-actions]

Skill Validation Report

Validating skill: /home/runner/work/icskills/icskills/skills/icp-cli

Structure

• Pass: SKILL.md found
• Pass: all files in references/ are referenced

Frontmatter

• Pass: name: "icp-cli" (valid)
• Pass: description: (626 chars)
• Pass: license: "Apache-2.0"
• Pass: metadata: (2 entries)

Tokens

• Warning: SKILL.md body is 7163 tokens (spec recommends < 5000)

Markdown

• Pass: no unclosed code fences found

Tokens

File	Tokens
SKILL.md body	7,163
references/binding-generation.md	1,031
references/dev-server.md	690
references/dfx-migration.md	2,620
Total	11,504

Content Analysis

Metric	Value
Word count	3,873
Code block ratio	0.17
Imperative ratio	0.12
Information density	0.14
Instruction specificity	0.80
Sections	20
List items	73
Code blocks	35

References Content Analysis

Metric	Value
Word count	2,133
Code block ratio	0.25
Imperative ratio	0.13
Information density	0.19
Instruction specificity	0.80
Sections	18
List items	33
Code blocks	12

Contamination Analysis

Metric	Value
Contamination level	low
Contamination score	0.12
Primary language category	shell
Scope breadth	3

• Warning: Language mismatch: config, javascript (2 categories differ from primary)

References Contamination Analysis

Metric	Value
Contamination level	low
Contamination score	0.03
Primary language category	javascript
Scope breadth	2

• Warning: Language mismatch: shell (1 category differ from primary)

Result: 1 warning

Project Checks

✓ Project checks passed for 1 skills (0 warnings)