ctf-ai-ml

Provides AI and machine learning techniques for CTF challenges. Use when attacking ML models, crafting adversarial examples, performing model extraction, prompt injection, membership inference, training data poisoning, fine-tuning manipulation, neural network analysis, LoRA adapter exploitation, LLM jailbreaking, or solving AI-related puzzles.

Quality

90%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Critical

Do not install without reviewing

Quality

Content

80%Weight 40%Scale 1-3

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

A concise, actionable quick-reference with executable code and inline formulas, weakened by missing referenced bundle files and the absence of validation checkpoints in its workflows.

Suggestions

Add the referenced bundle files (model-attacks.md, adversarial-ml.md, llm-attacks.md) or remove the links, since the one-level-deep progressive-disclosure structure currently points to non-existent files.

Add a brief validate-then-iterate checkpoint for the gradient-based and weight-perturbation techniques (e.g., confirm loss/weight-delta converged before declaring success) to raise workflow clarity.

For each technique entry, pair the one-liner with a minimal verification step so the catalog also functions as a lightweight workflow rather than only a lookup.

Dimension	Reasoning	Score
Conciseness	Lean reference where each technique is a one-liner with an inline formula or command and a pointer; it assumes Claude's competence and does not explain what neural networks or libraries are.	3 / 3
Actionability	Quick-start commands are executable (torch.load, safetensors inspection, curl prompt-injection probe) and inline formulas like `2W_orig - W_chal` and `x_adv = x + eps sign(grad_x(loss))` are copy-paste ready.	3 / 3
Workflow Clarity	Organized as a catalog with a useful "When to Pivot" decision section, but there is no sequenced multi-step workflow and no validation/verification checkpoints for the iterative gradient or weight-perturbation operations.	2 / 3
Progressive Disclosure	Structure is well-signaled with one-level-deep references and anchor links, but the referenced files (model-attacks.md, adversarial-ml.md, llm-attacks.md) do not exist in the bundle, so the intended navigation is broken.	2 / 3
	Total	10 / 12 Passed

Description

100%Weight 40%Scale 1-3

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, third-person description that states concrete capabilities and gives an explicit "Use when" trigger clause with good keyword coverage for the AI/ML CTF niche. No fluff or over-claims.

Dimension	Reasoning	Score
Specificity	Lists many concrete actions — "attacking ML models, crafting adversarial examples, performing model extraction, prompt injection, ... LoRA adapter exploitation, LLM jailbreaking" — rather than vague language.	3 / 3
Completeness	Explicitly answers what ("Provides AI and machine learning techniques for CTF challenges") and when ("Use when attacking ML models ... or solving AI-related puzzles") with an explicit trigger clause.	3 / 3
Trigger Term Quality	Good coverage of terms a CTF player would naturally say ("adversarial examples", "prompt injection", "LLM jailbreaking", "model extraction"); technical terms like "membership inference" fit the target audience.	3 / 3
Distinctiveness Conflict Risk	Clear AI/ML-CTF niche with distinct triggers unlikely to fire for crypto, reverse-engineering, or general skills; the "When to Pivot" body content reinforces separation.	3 / 3
	Total	12 / 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
relative_links	Relative link issues: 19 missing	Warning

	Total	14 / 16 Passed

Repository: ljagiello/ctf-skills
Path: ctf-ai-ml/SKILL.md
Commit: d19f35f

Reviewed: 9 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.