gan-style-harness

GAN-inspired Generator-Evaluator agent harness for building high-quality applications autonomously. Based on Anthropic's March 2026 harness design paper.

Quality

48%

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

Fix and improve this skill with Tessl

tessl review fix ./skills/gan-style-harness/SKILL.md

The canonical home for this skill is tdg-personal/gan-style-harness

Quality

Content

62%

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

The body is a well-sequenced, operationally rich harness guide with strong feedback-loop and validation design. Its weaknesses are conceptual padding and references to bundle files that are not actually present.

Suggestions

Either provide the referenced `scripts/gan-harness.sh`, `PLANNER_PROMPT.md`, and `EVALUATOR_PROMPT.md` as real bundle files, or remove/inline the references so the workflow is fully executable.

Trim the 'Core Insight' quote and GAN analogy to a one-line framing; Claude already understands the adversarial-loop concept.

Consider moving the evaluation rubric and environment-variable table into a reference file to reduce inline length and improve progressive disclosure.

Dimension	Reasoning	Score
Conciseness	Most content is operational (env tables, commands, anti-patterns), but the 'Core Insight' quote and extended GAN analogy explain concepts Claude largely knows and could be tightened, fitting the score-2 anchor.	2 / 3
Actionability	Concrete shell and `claude -p` commands are provided, but the manual workflow references `PLANNER_PROMPT.md`, `EVALUATOR_PROMPT.md`, and `scripts/gan-harness.sh` which do not exist as bundle files, leaving key details missing.	2 / 3
Workflow Clarity	The generator-evaluator loop is clearly sequenced (Steps 1-4, 'Repeat until pass threshold met') with explicit validation checkpoints: pass threshold, GAN_MAX_ITERATIONS, and the plateau-stop rule in anti-pattern #3.	3 / 3
Progressive Disclosure	Sections are well-organized, but everything sits inline in one ~270-line file and references to scripts/prompt files point to non-existent bundle files, matching the score-2 'content that should be separate is inline' anchor.	2 / 3
	Total	9 / 12 Passed

Description

35%

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

The description conveys the skill's niche and mechanism but is jargon-heavy and lacks explicit 'Use when...' trigger guidance. Natural user-facing trigger terms are absent, capping its discoverability.

Suggestions

Add an explicit 'Use when...' clause with natural trigger phrases a user would say, e.g. 'Use when building a complete app from a one-line prompt or when AI-slop aesthetics are unacceptable.'

Replace abstract framing ('GAN-inspired Generator-Evaluator agent harness') with concrete actions the skill performs, such as 'Plans, builds, and iteratively evaluates full-stack applications via a generator-evaluator feedback loop.'

Include common user terms like 'build an app from a prompt', 'production-quality app', or 'avoid AI slop' to improve trigger-term coverage.

Dimension	Reasoning	Score
Specificity	Names the domain ("Generator-Evaluator agent harness") and a broad action ("building high-quality applications autonomously"), but does not list multiple specific concrete actions like the score-3 anchor.	2 / 3
Completeness	It states what the skill does but omits any 'Use when...' trigger guidance, so per the judging guideline completeness is capped at 2.	2 / 3
Trigger Term Quality	The phrasing is technical jargon ("GAN-inspired Generator-Evaluator agent harness") with no natural trigger terms a user would actually say, matching the score-1 'technical jargon' anchor rather than score-2's 'some relevant keywords'.	1 / 3
Distinctiveness Conflict Risk	The GAN/generator-evaluator framing is a fairly distinct niche, but without explicit distinct triggers it could still overlap with general app-building skills, fitting the score-2 anchor.	2 / 3
	Total	7 / 12 Passed

Validation

87%

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 — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
metadata_version	'metadata.version' is missing	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	14 / 16 Passed

Repository: affaan-m/ECC
Path: skills/gan-style-harness/SKILL.md
Commit: 754b8dd

Reviewed: 2 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.