agentleFS
MysterionRiseMysterionRise/ctf-kit/skills/team-solve/SKILL.md

team-solve

Skill12 starsChanged 9 months agoMITView on GitHub

Orchestrate a multi-agent team to solve a CTF challenge. Use when a challenge is complex enough to benefit from parallel investigation — multiple possible approaches, layered encoding, or unclear category. Spawns 3 specialized teammates based on triage results. Each teammate gets a focused role within the detected category. Requires agent teams enabled (CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1). Triggers: "team solve", "parallel solve", "spawn team", "multi-agent", "use a team for this".

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 6 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Do NOT use when:

    line 28
  2. Distinctive:

    Run the triage pipeline first to determine the category:

    line 42
  3. Very specific:

    Read the JSON output. Key fields: category, confidence, secondary_categories.

    line 55
  4. Distinctive:

    Create an agent team with 3 teammates.

    line 63
  5. Distinctive:

    Let teammates self-claim tasks: after their initial assignment

    line 100
  6. Distinctive:

    Use when triage confidence is below 70% or multiple categories are detected.

    line 180
  7. Distinctive:

    All teammates get these automatically:

    line 209

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

CTF Team Solve

Orchestrate a team of 3 specialized teammates to solve a CTF challenge in parallel.

When to Use

Use this when:

  • A challenge is complex or multi-layered
  • Initial triage shows mixed signals or low confidence
  • You want to try multiple approaches simultaneously
  • Time pressure demands parallel investigation
  • The challenge has multiple files or stages

Do NOT use when:

  • The challenge is straightforward (single encoding, obvious category)
  • You already know exactly what tool to run
  • Agent teams are not enabled

Prerequisites

Agent teams must be enabled. Verify CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 is set in .claude/settings.json or environment.

Instructions

Step 1: Triage the Challenge

Run the triage pipeline first to determine the category:

bash scripts/triage-for-team.sh $ARGUMENTS

Or use the analyze skill's triage:

# From the ctf-kit plugin
bash ../analyze/scripts/triage.sh $ARGUMENTS

Read the JSON output. Key fields: category, confidence, secondary_categories.

Step 2: Decide Team Composition

Based on triage results, select the appropriate 3-teammate composition from the category tables below. If triage shows mixed signals (confidence < 70% or multiple secondary categories), use the Cross-Category Team composition.

Step 3: Spawn the Team

Create an agent team with 3 teammates. For each teammate:

  1. Name them by role (e.g., "rsa-specialist", "hash-cracker")
  2. Set the prompt using the role description and the challenge context
  3. Set plan approval for teammates using offensive tools (web, pwn, osint)
  4. Create initial tasks in the shared task list — one per teammate

Script paths in this skill are relative to its base directory. Teammates do not load this skill, so resolve every script to an absolute path before putting it in a spawn prompt.

Example spawn prompt to the lead:

Create an agent team with 3 teammates to solve this crypto challenge.

Teammate 1 — "classical-analyst":
  Focus on classical ciphers and encoding chains.
  Tools: frequency analysis, dcode.fr patterns, CyberChef recipes.
  Start by running: bash ../crypto/scripts/identify-hash.sh and bash ../misc/scripts/run-decode.sh

Teammate 2 — "rsa-specialist":
  Focus on RSA and asymmetric crypto attacks.
  Tools: RsaCtfTool, openssl, sage/python for math.
  Start by extracting any RSA parameters (n, e, c, p, q).

Teammate 3 — "symmetric-cracker":
  Focus on XOR, AES, and hash cracking.
  Tools: xortool, hashcat, john.
  Start by running: bash ../crypto/scripts/run-xortool.sh on any binary files.

Each teammate should broadcast findings to others when they discover something
that changes the understanding of the challenge. If a teammate finds a flag,
broadcast immediately and all teammates should verify.

Step 4: Coordinate

  • Let teammates self-claim tasks after their initial assignment
  • Monitor progress — check in if a teammate is stuck for more than a few minutes
  • Broadcast discoveries — tell a teammate to broadcast when they find something that shifts the approach
  • Synthesize — once 2+ teammates have findings, combine and identify the solution path
  • Shut down early — if one teammate solves it, shut down the others

Team Compositions by Category

Crypto Team

RoleNameFocusFirst Action
Classical & Encodingclassical-analystFrequency analysis, substitution ciphers, Vigenere, encoding chains, CyberChefRun ../misc/scripts/run-decode.sh and ../crypto/scripts/identify-hash.sh on all challenge files
Asymmetric & Mathrsa-specialistRSA factoring, padding oracles, Wiener/Boneh-Durfee, ECC, Diffie-HellmanExtract RSA parameters, run RsaCtfTool, try sage/python math attacks
Symmetric & Hashsymmetric-crackerXOR key recovery, AES mode attacks, hash cracking with hashcat/johnRun ../crypto/scripts/run-xortool.sh, identify and crack any hashes

Forensics Team

RoleNameFocusFirst Action
File & Diskfile-carverbinwalk, foremost, file carving, disk image mounting, deleted file recoveryRun ../forensics/scripts/run-binwalk.sh and ../forensics/scripts/extract-and-analyze.sh
Memorymemory-analystvolatility3 plugins, process trees, registry, command history, malware detectionRun ../forensics/scripts/run-volatility.sh with pslist, netscan, cmdline
Networknetwork-analysttshark, protocol analysis, stream extraction, DNS exfil detectionRun ../forensics/scripts/run-tshark.sh, extract HTTP objects and DNS queries

Web Team (plan approval required for Tier 2 tools)

RoleNameFocusFirst Action
Recon & Enumerationweb-reconDirectory scanning, technology fingerprinting, hidden paths, robots.txtRun ../web/scripts/run-gobuster.sh, check robots.txt, view source
Injectioninjection-testerSQLi, XSS, SSTI, command injection, path traversalTest input fields with common payloads, run sqlmap on parameters
Auth & Logicauth-analystJWT attacks, session handling, IDOR, privilege escalation, API abuseInspect cookies/tokens, test auth bypass, check API endpoints

Important: Spawn web teammates with plan approval required. The lead must review before any teammate sends requests to a target.

Pwn Team (plan approval required for remote connections)

RoleNameFocusFirst Action
Static Analystbinary-analystchecksec, disassembly, vulnerability ID, function mappingRun ../pwn/scripts/run-checksec.sh, map interesting functions
Exploit Developerexploit-devPayload crafting, ROP chains, shellcode, ret2libc, format stringsBuild exploit based on static analyst's findings
Dynamic Analystdynamic-analystGDB debugging, leak finding, heap state, ASLR bypassRun binary in debugger, find offsets, verify leaks

Important: Spawn the exploit-dev teammate with plan approval required before connecting to remote targets.

Reverse Team

RoleNameFocusFirst Action
Static Analyststatic-reverserradare2, Ghidra, decompilation, function listing, string analysisRun ../reverse/scripts/run-radare2.sh, identify interesting functions
Dynamic Analystdynamic-reverserltrace, strace, GDB, breakpoints, anti-debug bypass, runtime behaviorRun binary with ltrace/strace, set breakpoints on check functions
Algorithm Solveralgo-solverKeygen writing, constraint solving (z3), algorithm reimplementationReimplement validation logic, write solver script

Stego Team

RoleNameFocusFirst Action
Image Analystimage-analystzsteg LSB, steghide extraction, visual inspection, color plane analysisRun ../stego/scripts/run-zsteg.sh and ../stego/scripts/run-steghide.sh
Metadata & Structuremetadata-analystexiftool, binwalk appended data, file structure, IHDR manipulationRun ../stego/scripts/run-exiftool.sh and ../forensics/scripts/run-binwalk.sh
Audio & Advancedaudio-analystSpectrogram messages, audio LSB, DTMF tones, video frame analysisOpen in Audacity spectrogram view, check audio LSB encoding

OSINT Team (plan approval for external queries)

RoleNameFocusFirst Action
People & Socialsocial-investigatorSherlock username enum, social media profiling, identity correlationRun ../osint/scripts/run-sherlock.sh, cross-reference platforms
Domain & Infradomain-analystwhois, DNS records, subdomain enum, theHarvester, IP geolocationRun whois, dig, theHarvester on target domains
Geo & Mediageo-analystEXIF GPS extraction, reverse image search, visual landmark IDRun ../osint/scripts/run-exiftool.sh, identify visual clues

Important: Spawn OSINT teammates with plan approval required before hitting external services.

Misc Team

RoleNameFocusFirst Action
Encoding SpecialistdecoderCyberChef recipes, encoding chains, base conversions, custom encodingsRun ../misc/scripts/run-decode.sh, try multi-layer decoding
Esoteric & Codeesoteric-analystBrainfuck, Whitespace, JSFuck, custom languages, polyglotsIdentify language from character set, find interpreter
Puzzle & Logicpuzzle-solverPattern recognition, math puzzles, QR/barcodes, game theoryRun zbarimg on images, analyze patterns, solve constraints

Cross-Category Team (mixed signals or unclear category)

Use when triage confidence is below 70% or multiple categories are detected.

RoleNameFocusFirst Action
Analyst Aanalyst-aPrimary detected categoryUse the primary category's skill
Analyst Banalyst-bSecondary detected categoryUse the secondary category's skill
GeneralistgeneralistBroad investigation: strings, metadata, structure, encodingRun full triage, check all common patterns

Communication Patterns

When teammates should broadcast

  • Found a flag or flag fragment
  • Discovered a new file or layer (e.g., extracted an embedded file)
  • Identified the challenge type with high confidence
  • Hit a dead end — other teammates should deprioritize this approach
  • Found a password, key, or credential that others might need

When the lead should intervene

  • Two teammates are pursuing the same approach (redirect one)
  • A teammate has been stuck for >3 minutes
  • New information changes the strategy (reassign tasks)
  • One teammate solved it (shut down others)

Permissions Model

Tier 1 — Auto-allow (local analysis)

All teammates get these automatically:

file, strings, xxd, binwalk, foremost, exiftool, hashid, xortool,
zsteg, steghide, volatility3, tshark, checksec, ropgadget, radare2,
python3, zbarimg, ctf (CLI)

Tier 2 — Plan approval (connects to targets)

Teammates using these tools must have plan approval enabled:

sqlmap, gobuster, ffuf, nikto, pwntools (remote), nmap,
sherlock, theHarvester, curl/wget to targets

The lead reviews the plan before the teammate executes.

Tier 3 — User confirmation (resource-heavy or destructive)

These should prompt the user:

hashcat --gpu (resource-intensive), any write to target systems

Example Full Workflow

User: /ctf-kit:team-solve challenge.zip

Lead:
1. Runs triage → category=crypto, confidence=65%, secondary=stego
2. Low confidence → uses Cross-Category Team
3. Spawns:
   - "crypto-analyst" with crypto skill focus
   - "stego-analyst" with stego skill focus
   - "generalist" running broad analysis
4. Creates tasks:
   - Task 1: Crypto analysis of challenge files (crypto-analyst)
   - Task 2: Stego analysis of any images found (stego-analyst)
   - Task 3: Deep file analysis and strings search (generalist)
5. Generalist finds hidden PNG inside challenge file → broadcasts
6. Stego-analyst claims the PNG → finds LSB-encoded text
7. LSB text is a key → crypto-analyst uses it to decrypt
8. Flag found → broadcast → team shutdown

Related Skills

  • /ctf-kit:analyze — Triage a challenge (always run first)
  • /ctf-kit:compete — Competition mode (multiple challenges)
  • /ctf-kit:crypto through /ctf-kit:misc — Individual category skills

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.