ax-cpp-refine

Use when writing C++ code with `axllm` for reward-scored generation, iterative candidate improvement, evaluator feedback, and optimizer-backed refinement patterns.

Quality

83%

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

72%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.

A lean, well-organized reference card that respects the token budget and names a concrete API surface. It is weakened by the absence of an explicit workflow with validation checkpoints and by references to files (API.md, examples/) that are neither bundled nor located.

Suggestions

Add an explicit ordered workflow with a validation checkpoint, e.g.: 1. Pick the optimizer from the API surface; 2. Adapt the matching example from `examples/`; 3. Verify with the `no-key` scripted transport; 4. Only then run against a real provider.

State where the referenced artifacts (`API.md`, `axir-api.json`, `axir-capabilities.json`, `examples/`) live — a path relative to the skill or the `axllm` package — or bundle them, since none resolve inside the skill directory.

Make the Core Pattern snippet runnable (includes and a concrete instantiation) or explicitly label its placeholders and point to the exact example file that contains the full version.

Dimension	Reasoning	Score
Conciseness	The body is ~45 lean lines with no explanations of concepts Claude already knows — every section (Package Facts, Relevant API Surface, Guardrails) carries only facts the agent could not infer. Not 4: there is no over-explanation or padding to trim.	5 / 5
Actionability	The Core Pattern ("axllm::AxGEPA engine(reflection_client, options); auto result = engine.optimize(request, evaluator);") shows real syntax with named API symbols, and the API surface lists concrete symbol names. Not 5: the snippet uses undefined placeholders with no includes, so it is not copy-paste executable; not 3: it is genuine code shape with concrete symbols, not pseudocode.	4 / 5
Workflow Clarity	"When To Use" lists trigger conditions rather than steps, and the guardrails ("Start from package examples for exact native syntax", "Use `no-key` examples for deterministic local checks") serve only as implicit checkpoints — there is no explicit sequence or validation step for verifying generated code. Not 4: no ordered sequence with checkpoints is present; not 2: the guardrails do provide rough direction.	3 / 5
Progressive Disclosure	Sections are well organized, but the referenced artifacts ("API.md", "axir-api.json", "examples/") do not exist in the skill bundle and the body never states where they live, so the references are not clearly signaled or resolvable. Not 4: the reference navigation has real gaps; not 2: the inline structure itself is good and content is appropriately concise.	3 / 5
	Total	15 / 20 Passed

Description

87%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 description: it explicitly answers both what and when with concrete trigger phrases, uses specific capability language, and occupies a distinct niche. The only weaknesses are mild jargon ("reward-scored generation", "optimizer-backed refinement patterns") and missing natural synonyms.

Dimension	Reasoning	Score
Specificity	The description names the domain ("writing C++ code with `axllm`") and lists several concrete capability phrases ("reward-scored generation, iterative candidate improvement, evaluator feedback, and optimizer-backed refinement patterns"), matching the several-specific-actions anchor. It falls short of 5 because these are pattern labels rather than a comprehensive set of concrete actions.	4 / 5
Completeness	Both what and when are explicit: the capability list states what the skill covers, and "Use when writing C++ code with `axllm` for ..." is a concrete, specific trigger clause. Not 4: the anchor-4 gap ("'when' could be more explicit or specific") does not apply — the when clause is already fully explicit and paired with concrete triggers.	5 / 5
Trigger Term Quality	Natural domain terms are present — "C++", "axllm", "evaluator feedback", "optimizer", "refinement", "improvement" — giving good keyword coverage with only a few natural variations missing. Not 5: no synonyms or extension-style triggers, and phrases like "reward-scored generation" are internal jargon a user would rarely say.	4 / 5
Distinctiveness Conflict Risk	The combination of C++, `axllm`, and optimizer/evaluator terminology carves out a clear niche with distinct triggers and minimal conflict risk against generic coding or refinement skills.	5 / 5
	Total	18 / 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
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	15 / 16 Passed

Repository: ax-llm/ax
Path: packages/cpp/skills/ax-cpp-refine/SKILL.md
Commit: 3f630a0

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.