Solve ChallengeSAFE

Agent skills for solving CTF challenges - web exploitation, binary pwn, crypto, reverse engineering, forensics, OSINT, and more

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: solve-challenge
description: Solves CTF challenges by performing first-pass triage, identifying the dominant category, and routing execution to the right specialized ctf-* skill. Use when the user gives you a challenge bundle, a remote service, a suspicious file, or only a vague challenge description and you must determine where to start. Do not use it when the category is already clear and a specialized skill can be invoked directly; this is the dispatcher and recon entrypoint, not the deepest reference for category-specific techniques.
license: MIT
compatibility: Requires filesystem-based agent (Claude Code or similar) with bash, Python 3, and internet access. Orchestrates other ctf-* skills.
allowed-tools: Bash Read Write Edit Glob Grep Task WebFetch WebSearch Skill
metadata:
  user-invocable: "true"
  argument-hint: "[category] [challenge-file-or-url]"
---

# CTF Challenge Solver

You're a skilled CTF player. Your goal is to solve the challenge and find the flag.

## Environment Setup

Two setup strategies depending on your workflow:

### Pre-install (recommended before competitions)

Use the central installer entrypoint:

```bash
bash scripts/install_ctf_tools.sh all
```

Run a narrower mode when you only want one tool group:

```bash
bash scripts/install_ctf_tools.sh python
bash scripts/install_ctf_tools.sh apt
bash scripts/install_ctf_tools.sh brew
bash scripts/install_ctf_tools.sh gems
bash scripts/install_ctf_tools.sh go
bash scripts/install_ctf_tools.sh manual
```

The full package lists now live in [scripts/install_ctf_tools.sh](../scripts/install_ctf_tools.sh).

### On-demand (during challenges)

Each category skill's `SKILL.md` has a **Prerequisites** section listing only the tools needed for that category. Install as you go.

## Workflow

### Step 0: CTFd Platform Detection

If the CTF platform URL is known, check if it runs CTFd and switch to API-driven navigation:

```bash
# Detect CTFd (look for /api/v1/ and /themes/core/)
curl -s "$CTF_URL/api/v1/" | head -5
curl -s "$CTF_URL" | grep -oE '/themes/core/'
```

If CTFd is detected, **ask the user for their API token** (generated from CTFd Settings > Access Tokens). The token is not provided by default — the user must create one in the CTFd web UI first. Once provided, set the environment variables and proceed via API:

```bash
export CTF_URL="https://ctf.example.com"
export CTF_TOKEN="ctfd_..."  # Ask user for this
```

Invoke `/ctf-misc` and load its `ctfd-navigation.md` for the full API reference and Python client class.

### Step 1: Recon

1. **Explore files** -- List the challenge directory, run `file *` on everything
2. **Triage binaries** -- `strings`, `xxd | head`, `binwalk`, `checksec` on binaries
3. **Fetch links** -- If the challenge mentions URLs, fetch them FIRST for context
4. **Connect** -- Try remote services (`nc`) to understand what they expect
5. **Read hints** -- Challenge descriptions, filenames, and comments often contain clues

### Step 2: Categorize

Determine the primary category, then invoke the matching skill.

**By file type:**
- `.pcap`, `.pcapng`, `.evtx`, `.raw`, `.dd`, `.E01` -> forensics
- `.elf`, `.exe`, `.so`, `.dll`, binary with no extension -> reverse or pwn (check if remote service provided -- if yes, likely pwn)
- `.py`, `.sage`, `.txt` with numbers -> crypto
- `.apk`, `.wasm`, `.pyc` -> reverse
- `.safetensors`, `.pt`, `.pth`, `.bin`, `.onnx` -> ai-ml
- Web URL or source code with HTML/JS/PHP/templates -> web
- Images, audio, PDFs with no obvious content -> forensics (steganography)

**By challenge description keywords:**
- "buffer overflow", "ROP", "shellcode", "libc", "heap" -> pwn
- "RSA", "AES", "cipher", "encrypt", "prime", "modulus", "lattice", "LWE", "GCM" -> crypto
- "XSS", "SQL", "injection", "cookie", "JWT", "SSRF" -> web
- "disk image", "memory dump", "packet capture", "registry", "power trace", "side-channel", "spectrogram", "audio tracks", "MKV" -> forensics
- "find", "locate", "identify", "who", "where" -> osint
- "obfuscated", "packed", "C2", "malware", "beacon" -> malware
- "prompt injection", "LoRA", "adversarial", "model weights", "embedding" -> ai-ml
- "jail", "sandbox", "escape", "encoding", "signal", "game", "Nim", "commitment", "Gray code" -> misc

**By service behavior:**
- Port with interactive prompt, crash on long input -> pwn
- HTTP service -> web
- netcat with math/crypto puzzles -> crypto
- netcat with restricted shell or eval -> misc (jail)

### Step 3: Invoke the Category Skill

Once you identify the category, **invoke the matching skill** to get specialized techniques:

| Category | Invoke | When to Use |
|----------|--------|-------------|
| Web | `/ctf-web` | XSS, SQLi, SSTI, SSRF, JWT, file uploads, prototype pollution |
| Pwn | `/ctf-pwn` | Buffer overflow, format string, heap, ROP, sandbox escape |
| Crypto | `/ctf-crypto` | RSA, AES, ECC, PRNG, ZKP, classical ciphers |
| Reverse | `/ctf-reverse` | Binary analysis, game clients, VMs, obfuscated code |
| Forensics | `/ctf-forensics` | Disk images, memory dumps, event logs, stego, network captures |
| OSINT | `/ctf-osint` | Social media, geolocation, DNS, public records |
| Malware | `/ctf-malware` | Obfuscated scripts, C2 traffic, PE/.NET analysis |
| AI/ML | `/ctf-ai-ml` | Model weights (`.safetensors`, `.pt`, `.pth`, `.bin`, `.onnx`), prompt injection, LoRA adapters, adversarial examples |
| Misc | `/ctf-misc` | Jails, encodings, RF/SDR, esoteric languages, constraint solving |

You can also invoke `/ctf-<category>` to load the full skill instructions with detailed techniques.

### Step 4: Pivot When Stuck

If your first approach doesn't work:

1. **Re-examine assumptions** -- Is this really the category you think? A "web" challenge might need crypto for JWT forgery. A "forensics" PCAP might contain a pwn exploit to replay.
2. **Try a different category skill** -- Many challenges span multiple categories. Invoke a second skill for the cross-cutting technique.
3. **Look for what you misse

SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

What the source does

Filesystem

none-observed

Network

none-observed

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (0)

No findings outside the package's declared scope.

Gates applied: no_behavioural_pass.