AGENT SKILL

nice-try-gpt

Analyzes authorized CTF challenges by reproducing intended solves, identifying cheap LLM shortcuts, applying minimal human-friendly transformations, and verifying results end-to-end. Appropriate when reviewing or adapting CTFs to reduce pattern-matching shortcuts without materially increasing human difficulty.

License 许可
GPL-3.0
所属仓库
bg-szy/TOP-SKILLS
最近更新
2026-11-15
Stars
★ 6
Forks
0

01 前言 Front Matter

name
nice-try-gpt
license
GPL-3.0-only
metadata
{ "author": "aleff-github", "version": "0.2.0" }
description
Analyzes authorized CTF challenges by reproducing intended solves, identifying cheap LLM shortcuts, applying minimal human-friendly transformations, and verifying results end-to-end. Appropriate when reviewing or adapting CTFs to reduce pattern-matching shortcuts without materially increasing human difficulty.

02 资源文件 1

03 SKILL.md 正文

NiceTryGPT

Less pattern matching. More actual hacking.

Use this skill only for CTFs, training labs, and systems the user is authorized to test.

Core contract

The goal is not to make the challenge harder in general. The goal is to reduce one cheap LLM shortcut while keeping the challenge fair and recognizable to a human player.

Preserve these invariants:

  1. Same intended vulnerability class.
  2. Same learning objective.
  3. Same prerequisite knowledge.
  4. Same flag or success semantics.
  5. Roughly the same human difficulty band.

Default to one resistance change. Use a second only when the first is insufficient and the Human Cost Gate still passes.

It is valid to make no change.

Use these final statuses exactly:

  • BASELINE FAILED
  • NO CHANGE NEEDED
  • TRANSFORMED PASS
  • TRANSFORMED FAILED

1. Understand

Read only what is needed to understand and run the challenge:

  • player-facing description;
  • entry point;
  • relevant source/configuration;
  • run or build instructions;
  • flag format or equivalent success condition;
  • intended learning objective.

Before editing, record:

  • vulnerability class;
  • intended solve path;
  • expected player knowledge;
  • approximate difficulty band;
  • clean-start procedure.

Do not modify files yet.

2. Baseline solve

Start from a clean state and solve the original challenge end-to-end through the player-facing surface.

A valid baseline must obtain the runtime flag or equivalent success condition through the intended vulnerability.

Do not count any of these as a solve:

  • reading the flag directly from source;
  • reading fixture or environment values directly;
  • inspecting container internals only to extract the answer;
  • assuming the documented solve works without reproducing it.

If the baseline cannot be reproduced, stop with BASELINE FAILED. Do not transform an unverified challenge.

3. Identify the cheapest shortcut

Ask:

What lets an LLM jump from observation to solution without enough interaction or hypothesis testing?

Choose one primary shortcut, supported by evidence from the baseline. Examples:

  • a parameter or endpoint practically names the vulnerability;
  • the next obvious numeric ID is the target;
  • a textbook payload works immediately with no discovery;
  • all solve-relevant facts appear in one static response;
  • an error message reveals the intended path;
  • a static value removes the need to observe runtime behavior.

Do not invent a problem just to justify a transformation.

4. Pick the smallest resistance move

Choose zero, one, or at most two lightweight moves:

  • Pattern break — remove an overly explicit cue without hiding the vulnerability.
  • Runtime discovery — move one solve-relevant fact into normal runtime behavior.
  • Context split — require connecting two nearby pieces of ordinary application behavior.
  • State dependency — let a small amount of ordinary state matter.
  • Semantic decoy — add one plausible but cheaply falsifiable attack surface.

Read references/resistance-patterns.md only when selection guidance is useful.

Do not stack patterns by default.

5. Human Cost Gate

Reject a proposal unless all of these remain true:

  • same vulnerability class;
  • same learning objective;
  • same prerequisite knowledge;
  • no new exploit primitive required;
  • no brute force;
  • no CAPTCHA or human-verification gimmick;
  • no huge context, token flooding, or pointless encoding;
  • no obscure external trivia;
  • no artificial multi-stage chain;
  • normally no more than 1–3 additional meaningful player actions;
  • expected solve difficulty stays in the same band.

A meaningful action is an interaction that materially advances or rejects a hypothesis, such as one request, command, or inspection step.

For semantic decoys:

  • default to zero;
  • if useful, add only one semantic decoy;
  • make it dismissible in 1–2 normal interactions;
  • never use fake flags, destructive traps, or punishment.

If a useful resistance change fails this gate, keep the original challenge and report NO CHANGE NEEDED.

6. Modify

Implement the smallest diff that satisfies the chosen strategy.

Preserve normal application behavior. Avoid cosmetic rewrites that do not change the shortcut.

Runtime randomization is acceptable only when:

  • the player can discover the randomized value through normal interaction;
  • it never requires guessing or brute force;
  • it does not make the challenge flaky.

Do not hide the flag behind randomness or instability.

7. Verify end-to-end

Restart or rebuild from a clean state.

Verify all applicable checks:

  1. application starts normally;
  2. legitimate player-facing functionality still works;
  3. the intended vulnerability still exists;
  4. the original cheap shortcut is actually reduced or removed;
  5. the transformed solve works end-to-end;
  6. the runtime flag is obtained through the vulnerability;
  7. the flag is not leaked by unrelated surfaces;
  8. any decoy has a normal path that works and an obvious exploit probe that fails safely.

If verification fails, simplify or revert the transformation. End with TRANSFORMED FAILED if a valid transformed challenge cannot be restored.

8. Fresh-solver check

When a genuinely isolated context, subagent, or separate model is available, give it only the player-facing challenge information and ask it to solve the transformed challenge.

Do not reveal:

  • the baseline solution;
  • the shortcut analysis;
  • the transformation plan.

If no independent solver is available, mark the result NOT TESTED.

Never claim that self-review proves LLM resistance.

9. Report

Create or update nicetrygpt-report.md in the challenge root.

Keep it short and include:

  • final status;
  • baseline result;
  • vulnerability class;
  • shortcut identified;
  • transformation used, or NO CHANGE NEEDED;
  • human-cost estimate in additional meaningful actions;
  • original difficulty band;
  • post-change difficulty band;
  • post-change E2E result;
  • fresh-solver result or NOT TESTED;
  • files changed.

Success rule

A transformation succeeds only when the learning objective is preserved, the challenge still works end-to-end, and the identified shortcut is reduced without materially worsening the human experience.

If those goals conflict, favor the human player.

04 同仓库更多 Skills

bg-szy
bg-szy/TOP-SKILLS

serpmaven

Scrape Google search results (SERP) to JSON with SERP Maven (serpmaven) - a real Chrome browser on a residential proxy with sticky sessions. Run it as a CLI, a Python library, a local HTTP search API (`serve`), an MCP server or a Docker image; set the proxy login, workers, pages and cost tracking. Use when the user wants Google results, rank tracking, a self-hosted SERP API, or live Google search for an AI agent.

★ 6
bg-szy
bg-szy/TOP-SKILLS

revnu-landing-page-audit

Find out why a landing page isn't turning visitors into signups, demos or sales. Audits one page against what the ad, post or search that sends people there promised, then checks the first screen, proof, call to action, friction and mobile. Returns at most 7 prioritized fixes, plus what is fine and what could not be checked; rewrites specific lines and fixes the page in the repo when asked. Never promises a conversion lift. Works without a Revnu account. Use when the user says or asks: "audit my landing page", "why isn't this page converting", "review my homepage", "my ads get clicks but no signups", "people bounce from the pricing page", "CRO review", "improve conversion rate", "is my hero section clear", "roast my landing page", "rewrite my headline", "what's wrong with this page", "landing page feedback", "make the signup page convert better". Not for SEO or AI-search audits, page speed engineering on its own, or building a new page from scratch.

★ 6
bg-szy
bg-szy/TOP-SKILLS

revnu-growth-context

Write down once who buys, what they're stuck on, what it costs, what proof you have and what you've already tried, so every growth task starts from facts instead of a questionnaire. Creates or updates .agents/revnu-growth.md from the repo, the live site and the user, and imports an existing product-marketing.md. Use when the user says "set up our marketing context", "who is our ideal customer", "write our ICP", "describe the product for marketing", "positioning doc", "what's our brand voice", "remember this about our customers", "log what we tried", "what have we learned from marketing", "we changed pricing, update the context", "update our growth notes", "summarize the business for the agent", "import my product-marketing.md", "start a growth doc", or before a first marketing, SEO, outreach or launch task when no context file exists. Works without a Revnu account.

★ 6
bg-szy
bg-szy/TOP-SKILLS

revnu-get-listed

Get onto the third-party pages AI answers and search cite when buyers ask for the best tool in your category: "best X" roundups, directories, review sites, GitHub awesome lists, community threads and comparison pages. Finds which ones are cited, ranks them by fit and effort, and drafts each submission, PR or pitch by the target's own rules. Use when the user says "get us listed", "which directories should we submit to", "how do we get into best-of lists", "why does ChatGPT recommend competitors and not us", "get mentioned in AI answers", "submit to awesome lists", "get on G2", "where are our competitors listed", "pitch listicle authors", "find backlink opportunities", "launch directories", "get into roundups", "Product Hunt alternatives", "where should we be listed", "get cited by Perplexity". Never submits, pays or posts without a yes. Works without a Revnu account.

★ 6
bg-szy
bg-szy/TOP-SKILLS

revnu-ai-readiness-audit

Audit whether your website gives ChatGPT, Perplexity, Claude, Gemini and Google's AI Overviews something worth quoting, and fix the pages that don't. Use when the user says or asks "is my site ready for AI search", "audit my site for AI answers", "AEO audit", "GEO audit", "why doesn't ChatGPT quote our pricing", "make our pages citable by AI", "do I need llms.txt", "does schema help with ChatGPT", "check our structured data", "is our content visible without JavaScript", "AI SEO audit", "optimize our homepage for AI answers", "LLM SEO check", "why does AI get our product wrong", "fix our site for AI Overviews". Checks fetchability and indexability, answers in the raw HTML, entity facts, schema (including JS-injected markup), headings, dates and llms.txt, then gives at most 10 prioritised fixes and edits the repo when asked. Works without a Revnu account.

★ 6
bg-szy
bg-szy/TOP-SKILLS

proxy-troubleshooting

Diagnose residential proxy failures, with NodeMaven's gateway as the reference - 407 with correct credentials, 406, 410, 500, a connection that hangs, ERR_TUNNEL_CONNECTION_FAILED in a browser, Playwright or Puppeteer ignoring proxy credentials, a sticky session that changes IP, geo-targeting that is silently ignored, passwords with special characters, and unexpected traffic from headless Chrome. Use when a proxy connection fails, returns the wrong location, or costs more traffic than expected.

★ 6