CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

ax-go-refine

Use when writing Go code with `github.com/ax-llm/ax/packages/go` for reward-scored generation, iterative candidate improvement, evaluator feedback, and optimizer-backed refinement patterns.

Quality

75%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./packages/go/skills/ax-go-refine/SKILL.md
SKILL.md
Quality
Evals
Security

Quality

Content

68%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 notably concise, well-organized body that assumes Claude's competence and stays free of padding. Its weaknesses are executability — the core pattern is illustrative pseudocode with an inconsistent package prefix and no runnable example — and the absence of a sequenced workflow for choosing and applying an optimizer.

Suggestions

Replace or augment the 'Core Pattern' snippet with one complete, runnable example (imports, a concrete evaluator, a no-key transport setup) so the code is copy-paste ready.

Resolve the receiver naming inconsistency (`ax.NewGEPA` vs `axllm.AxGEPA`) and add a one-line rule for when to pick each optimizer (GEPA vs BootstrapFewShot vs Playbook).

Add a short ordered workflow with a verification step, e.g. 1) check examples/ for the native call shape, 2) run the no-key example deterministically, 3) only then wire a real provider.

DimensionReasoningScore

Conciseness

The body is lean and fact-dense: terse sections ('Package Facts', 'Relevant API Surface', 'Guardrails') with zero explanation of concepts Claude already knows, and every line adds package-specific information. It clearly matches the 'every token earns its place' anchor; the neighboring 4-level anchor would require some trimming, which is absent here.

5 / 5

Actionability

The 'Core Pattern' snippet is pseudocode rather than executable: it references undefined variables (`reflectionClient`, `request`, `evaluator`) and uses `ax.NewGEPA` while the API list names `axllm.AxGEPA`, an unresolved naming inconsistency. The API surface is a bare symbol list without signatures or a worked example, matching the 'some concrete guidance but incomplete; pseudocode instead of executable code' anchor rather than the mostly-executable level 4.

3 / 5

Workflow Clarity

There is no sequenced process for accomplishing a refinement task — no guidance on when to choose `AxBootstrapFewShot` vs `AxGEPA` vs `Playbook`, or how to go from evaluator definition to optimized result. The guardrails and 'When To Use' give direction without order or checkpoints, which fits the 'sequence present but checkpoints missing or implicit' band; it cannot score higher because the core action is not unambiguous enough for the simple-skill exception.

3 / 5

Progressive Disclosure

The skill is under 50 lines, cleanly sectioned, and its pointers to deeper material ('API.md', 'axir-api.json', 'axir-capabilities.json', 'examples/') are clearly signaled in 'Package Facts'. This lands on the 'good structure; references mostly clear; minor organization gaps' anchor rather than 5 because the referenced artifacts live in the target package rather than a bundled references directory, leaving navigation to the agent, and the inlined API symbol list could arguably sit in a separate reference file.

4 / 5

Total

15

/

20

Passed

Description

71%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 well-scoped, third-person description with an explicit 'Use when' trigger and a specific package boundary that makes it highly distinctive. Its main weakness is trigger-term phrasing that favors internal jargon over the natural words a user would say when needing this skill.

Suggestions

Add natural user-facing trigger phrases alongside the jargon, e.g. 'Use when improving LLM outputs in Go, applying GEPA or few-shot optimizers, or wiring evaluator feedback loops'.

Consider naming the concrete optimizer entry points users would mention (e.g. 'GEPA', 'bootstrap few-shot', 'playbook') so the description matches how the need is actually voiced.

Convert one or two capability noun phrases into verb+object actions (e.g. 'score candidates against evaluators and iterate on the best one') to sharpen the 'what' clause.

DimensionReasoningScore

Specificity

The description lists four named capability areas ('reward-scored generation, iterative candidate improvement, evaluator feedback, and optimizer-backed refinement patterns'), giving several specific actions with only minor coverage gaps. It falls short of a 5 because the items are pattern-category noun phrases rather than crisp, comprehensive verb+object actions like the top anchor's examples.

4 / 5

Completeness

It explicitly answers both parts: 'Use when writing Go code with `github.com/ax-llm/ax/packages/go`' gives a clear, package-scoped when, and the trailing clause states what the skill covers. The 'when' is concrete, but the 'what' is stated as capability patterns rather than explicit operations, so it sits at anchor 4 rather than the fully concrete trigger-phrase level of 5.

4 / 5

Trigger Term Quality

Some relevant keywords are present ('Go code', 'refinement', 'evaluator feedback') but the phrasing leans technical ('reward-scored generation', 'optimizer-backed refinement patterns') rather than natural user language, and common variations users would actually say ('improve outputs', 'optimize prompts', 'GEPA', 'few-shot') are missing. This matches the 'some relevant keywords but missing common variations or synonyms' anchor, not the 4-level's good coverage.

3 / 5

Distinctiveness Conflict Risk

The description is pinned to one specific package path (`github.com/ax-llm/ax/packages/go`) with distinct, domain-specific triggers, giving it a clear niche with minimal conflict risk against other skills. No neighboring anchor fits better, since even the 4-level anchor presumes some overlap risk that this scoping avoids.

5 / 5

Total

16

/

20

Passed

Validation

93%

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

CriteriaDescriptionResult

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/go/skills/ax-go-refine/SKILL.md

Reviewed

Table of Contents

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.