cursor-ask

Read-only delegate to cursor-agent (Composer) for questions, investigation, design discussion, and adversarial sanity checks. No worktree, no cherry-pick, no Lead diff review — cursor-agent is locked to ask mode and cannot write. Use when user says: ask cursor, cursor sanity check, get a second opinion, adversarial review, design discussion, investigate with cursor, cursor:ask. Do NOT load for: implementation, refactor, file edits, commit/push work, anything requiring write access (use cursor:do or breezing --cursor instead).

Quality

85%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Quality

Content

75%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body is highly actionable with a complete executable script and a clear sequenced workflow backed by validation guards and a trust-boundary checklist. Its main weakness is conciseness — the banner/narration contract is repeated across three sections — and some inline detail could be offloaded to referenced files.

Suggestions

Collapse the banner/plan narration into a single source of truth: define it once in Narration Rules and have Steps 0–1 reference it instead of restating it, to remove the triplicated explanation.

Move the verbose violation/normal narration examples and possibly the Trust Boundary precondition detail into the referenced `.claude/rules/cursor-cli-only.md` (or a references file), keeping SKILL.md as a lean overview.

Add an explicit error-recovery loop in Step 2 (e.g., on root-not-found or cursor-companion failure, surface the remediation step rather than only exiting) to push workflow clarity from 4 to 5.

Dimension	Reasoning	Score
Conciseness	Mostly efficient and project-specific (it does not explain generic concepts Claude already knows), but the banner/plan narration is restated across three sections — Narration Rules, Step 0, and Step 1 — and the violation/normal example block could be tightened, so it earns "mostly efficient but could be tightened" rather than the lean 4 or 5.	3 / 5
Actionability	Step 2 provides a fully executable, copy-paste-ready bash block with real error handling (empty-question exit, plugin-root resolution fallback, exit 2 on root-not-found), and Quick Reference plus the use-case table cover the common invocation cases.	5 / 5
Workflow Clarity	Steps 0–3 are clearly sequenced with explicit validation checkpoints (empty-question guard, root-resolution guard, a Step 3 summary checklist, and the Trust Boundary preconditions table), but it lacks a true validate→fix→retry feedback loop, so it sits at 4 rather than 5.	4 / 5
Progressive Disclosure	No bundle files exist, so scoring is against structure alone; the body is well-organized into clearly headed sections and the Related Skills / Rules section gives well-signaled one-level-deep references, but the inlined Narration Rules detail and Trust Boundary table could arguably live in the referenced rules file, leaving it just short of a 5.	4 / 5
	Total	16 / 20 Passed

Description

95%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong, well-targeted description: it states the read-only delegation purpose, lists concrete action types, supplies an extensive set of natural trigger phrases, and disambiguates from write-capable sibling skills via negative triggers. The only mild weakness is that the named actions are delegation categories rather than operationally concrete verbs.

Dimension	Reasoning	Score
Specificity	Names the domain (read-only cursor-agent delegation) and several specific action types — "questions, investigation, design discussion, and adversarial sanity checks" — plus the hard read-only constraint; not a 5 because the verbs are delegation-mode categories rather than operationally concrete operations like the rubric's 5-anchor examples.	4 / 5
Completeness	Explicitly answers both what ("Read-only delegate to cursor-agent ... locked to ask mode and cannot write") and when ("Use when user says: ..." with concrete triggers), and adds negative triggers, matching the 5-anchor requirement for concrete trigger phrases on both sides.	5 / 5
Trigger Term Quality	Lists comprehensive natural trigger phrases a user would actually say — "ask cursor", "cursor sanity check", "get a second opinion", "adversarial review", "design discussion", "investigate with cursor", "cursor:ask" — including the slash-command synonym form.	5 / 5
Distinctiveness Conflict Risk	Occupies a clear niche (read-only ask delegation) and explicitly disambiguates from sibling skills via "Do NOT load for: implementation, refactor, file edits ... use cursor:do or breezing --cursor instead", giving minimal conflict risk.	5 / 5
	Total	19 / 20 Passed

Validation

93%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
referenced_paths_exist	Referenced path issues: 3 missing	Warning

	Total	15 / 16 Passed

Repository: Chachamaru127/claude-code-harness
Path: opencode/skills/cursor-ask/SKILL.md
Commit: 2b2b748

Reviewed: 3 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.