chore: sync mops-cli skill from upstream cli-v2.15.2

[marc0olo]

Summary

Syncs the mops-cli skill from caffeinelabs/mops cli-v2.14.0 → cli-v2.15.2 (commit 564a380). Closes #234.

Per .claude/upstream.md, mops-cli has no icskills-owned sections — the body is 1:1 with upstream and only the frontmatter differs. Applied the full upstream delta and refreshed the frontmatter/tracking.

Upstream body changes applied

• New mops deployed command — post-deploy hook that promotes the built .most to the check-stable baseline (mops deployed init for bootstrap, mops deployed after each deploy).
• New mops generate candid command — (re)generates the curated .did from Motoko source.
• check-stable bootstrap rewritten — replaces the hand-written trivial .most snippet with mops deployed init; the [check-stable].path example is now deployed/backend.most, and the description explains it runs ICP's upgrade-time compatibility check locally.
• --no-check-limit flag documented for mops check / mops check-stable / mops lint, plus the new pending-migration reporting behavior.
• mops check --fix now notes read-only files (e.g. frozen migrations) are skipped with a warning.

Other

• Bumped compatibility to mops >= 2.15.2 (the new commands require it).
• Updated .claude/upstream.md — tag, full commit SHA, and last-synced date.
• Fixed a pre-existing local drift: the mops update --major / --patch ordering now matches upstream.
• Added eval cases covering the new commands and the changed check-stable bootstrap workflow (including an adversarial case guarding against the now-removed hand-written .most).

Validation

npm run validate — all 26 skills pass (warnings only, none new for mops-cli).

Eval results

Ran the new/changed cases with baseline. Every new command shows a clear with-skill vs baseline delta — without the skill, agents hand-write .most files, deny mops generate candid exists, and fall back to dfx generate.

Output evals (new cases, with vs without skill)

Case	WITH skill	WITHOUT skill
check-stable baseline bootstrap	3/4	1/4
Post-deploy .most promotion	4/4	3/4
Regenerating candid interface	4/4	0/4
Autofix trimmed migrations with --no-check-limit	3/3	0/3*

Notable baseline failures (without skill):

• check-stable bootstrap: suggested touch .baseline.most instead of mops deployed init.
• candid: claimed "mops doesn't have a command for generating .did files" and recommended dfx generate — the exact hallucination this skill prevents.

* The --no-check-limit baseline run hit the 120s harness timeout and returned empty, so 0/3 reflects no output rather than a wrong answer; the with-skill run passed 3/3.

The check-stable baseline bootstrap 3/4 miss is a soft "may note the default deployed/<name>.most path" behavior; the three hard behaviors passed.

Trigger evals

Should trigger:      14/14 correct  (includes the 2 new queries: candid regeneration, check-stable sync)
Should NOT trigger:   8/8  correct

🤖 Generated with Claude Code

[github-actions]

Skill Validation Report

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

Structure

• Pass: SKILL.md found

Frontmatter

• Pass: name: "mops-cli" (valid)
• Pass: description: (341 chars)
• Pass: license: "Apache-2.0"
• Pass: compatibility: (14 chars)
• Pass: metadata: (2 entries)

Markdown

• Pass: no unclosed code fences found

Tokens

File	Tokens
SKILL.md body	2,708
Total	2,708

Content Analysis

Metric	Value
Word count	1,486
Code block ratio	0.31
Imperative ratio	0.09
Information density	0.20
Instruction specificity	0.60
Sections	23
List items	9
Code blocks	15

Contamination Analysis

Metric	Value
Contamination level	low
Contamination score	0.01
Primary language category	shell
Scope breadth	2

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

Result: passed

Project Checks

✓ Project checks passed for 1 skills (0 warnings)