Sato Hub
Agent skills

Crypto Agent Skills

Browse skills for crypto agents and see what each one touches before you install it.

Browse skillsCheck a skill
  • Skills indexed565
  • With a flag235
  • Registries2
  • Updated2026-10-19

Browse skills

Download data
AllHas a flagNo flags

8 skills · as of 2026-10-19

FlagsFlagsA flag marks something the skill's own files do, like handling keys or running a shell command. It describes the skill and does not accuse it. A skill with no flags is not cleared either.

  • ctf-web

    skills.shOther

    Provides web exploitation techniques for CTF challenges. Do not use it for native binary memory corr.

    Installs
    9,395
    Hosts contacted
    30
    Handles keysHandles keysCreates or reads a private key or seed phrase. Normal for wallets.Asks for secretsAsks for secretsAsks you to paste a key, password or seed phrase.+4More flagsReads secret files · Runs fetched code · Registers agent · Runs on a schedule
    View skill

  • ctf-pwn

    skills.shOther

    Provides binary exploitation techniques for CTF challenges. Do not.

    Installs
    9,139
    Hosts contacted
    1
    Runs fetched codeRuns fetched codeRuns code it downloaded or decoded while running.
    View skill

  • solve-challenge

    skills.shOther

    Solves CTF challenges by performing first-pass triage, identifying the dominant category, and routing execution to the…

    Installs
    8,644
    Hosts contacted
    1
    Runs fetched codeRuns fetched codeRuns code it downloaded or decoded while running.
    View skill

  • ctf-misc

    skills.shOther

    Provides miscellaneous CTF challenge techniques for problems that do not cleanly fit the main categories.

    Installs
    8,574
    Hosts contacted
    14
    Handles keysHandles keysCreates or reads a private key or seed phrase. Normal for wallets.Runs fetched codeRuns fetched codeRuns code it downloaded or decoded while running.+1More flagsRuns on a schedule
    View skill

  • ctf-ai-ml

    skills.shOther

    Provides AI and machine learning techniques for CTF challenges.

    Installs
    7,461
    Hosts contacted
    0
    Runs fetched codeRuns fetched codeRuns code it downloaded or decoded while running.
    View skill

  • evaluation-methodology

    skills.shOther

    PluginEval quality methodology, covering dimensions, rubrics, and scoring formulas.

    Installs
    7,097
    Hosts contacted
    0
    Runs fetched codeRuns fetched codeRuns code it downloaded or decoded while running.
    View skill

  • sharp-edges

    skills.shOther

    Identifies error-prone APIs, dangerous configurations, and footgun designs that enable security mistakes.

    Installs
    6,640
    Hosts contacted
    3
    Runs fetched codeRuns fetched codeRuns code it downloaded or decoded while running.
    View skill

  • security-review

    skills.shOther

    Provides comprehensive security checklist and patterns.

    Installs
    5,629
    Hosts contacted
    4
    Reads secret filesReads secret filesReads files that hold secrets, like ~/.ssh or .env.Runs fetched codeRuns fetched codeRuns code it downloaded or decoded while running.
    View skill
How we read skillsShowHide

A skill is a document an agent follows. Before you install one that can touch a wallet, this page shows what it declares and what its text does: the hosts it contacts, whether it generates keys, whether it asks you to paste a credential, whether it pipes the internet into a shell.

Each disclosure is a static pattern scan of the skill's own files. It describes and does not judge. A flag is not an accusation: a wallet skill that generates keys is doing its job. No flag is not a clearance. Where a registry publishes its own scan result, the skill's page shows it as that registry's result, not ours.

We index crypto-related skills from ClawHub and skills.sh and scan them again every week. When a skill's SKILL.md file changes, we record the new version.

Categories come from the words in each skill's name and description. Each skill sits in one category.

  • Handles keys: Generates or handles private keys.
  • Asks for secrets: Asks the user to paste a credential.
  • Reads secret files: Reads secret paths (~/.ssh, .env …).
  • Pipes to shell: Pipes a remote script into a shell.
  • Runs fetched code: Executes fetched or decoded code.
  • All tools: Grants itself unrestricted tools.
  • Registers agent: Registers the agent with a third-party host.
  • Runs on a schedule: Schedules itself to run repeatedly.

Download the data as JSON (CC-BY 4.0).

Reference

QuestionsShowHide

Questions people ask about crypto agent skills

What is an agent skill?
A document an agent follows — instructions, often with scripts beside them — installed into an agent's context rather than called as a tool. Sato Hub indexes 565 crypto-relevant ones as of 2026-10-19 from 2 public registries and re-scans weekly, recording a changed file hash.
What is a disclosure, and what is it not?
A static pattern scan of the skill's own files: hosts it contacts, whether it generates keys, whether it asks for a credential, whether it pipes the internet into a shell, what tools it grants. It describes. It is never a verdict, and no flag is not a clearance.
Which disclosures come up most?
Generates or handles private keys (163), Schedules itself to run repeatedly (41), Reads secret paths (~/.ssh, .env …) (38), Executes fetched or decoded code (27), Grants itself unrestricted tools (27) as of 2026-10-19. 235 of 565 skills carry at least one. A flag is not an accusation — a wallet skill that generates keys is doing its job.
Does Sato Hub say a skill can be installed without risk?
No, and it never will. Where a registry publishes its own scan verdict it is shown as that registry's, attributed. What is published here is what the text of the skill does, so you can decide before it reaches a wallet.
SourcesShowHide

Sources — the skill format and the tool protocol

Cite this pageShowHide

Cite this page

Sato Hub. "Sato Hub crypto agent skills." Sato Hub, updated 2026-10-23, accessed 2026-10-23. https://satohub.ai/skills

Data last refreshed 2026-10-23; this page is rebuilt daily. Citations carry the date so a reader can tell which snapshot a claim came from. Catalog data is licensed CC-BY-4.0.