Trail of Bits Skills
trailofbits/skillsTrail of Bits' marketplace of 97 security skills: smart-contract scanners, C/C++ and Rust review, Semgrep and CodeQL workflows, variant analysis and audit helpers.
Recommended
Broad access Depth 4/5
ctf-skills Claude skills
Twelve technique-reference skills for solving capture-the-flag challenges: web, pwn, crypto, reverse engineering, forensics, OSINT, malware analysis and AI/ML.
Our verdict Recommended with caveats
Deep, well-organized CTF reference content; use it only on challenges and systems you are authorized to attack, and expect a heavy tool setup.
What ctf-skills does
ctf-skills gives an agent a large library of competition techniques organized by category. Each category skill (ctf-web, ctf-pwn, ctf-crypto, ctf-reverse, ctf-forensics, ctf-osint, ctf-malware, ctf-misc, ctf-ai-ml) has a short SKILL.md with a prerequisites list and one-line technique summaries, linking to many supporting files with worked details. Several category skills run to 28 to 42 KB of instructions, and the pack has about 200 reference files. Other skills handle the workflow: solve-challenge, ctf-writeup, and ctf-similar-search, which queries Tavily to find related past challenges.
A few small helper scripts exist (for example a concurrent fuzzer in ctf-web and exploit-template scripts in ctf-pwn), and the repo has 46 eval-related files. The README describes a central tool installer and an on-demand approach where each skill lists the tools it needs; the installer script it mentions was not present in the copy we reviewed. The material is offensive-security knowledge intended for competitions and authorized practice. Setup can pull in many system packages.
Good for
- CTF players using an agent as a technique lookup and solver
- Security students practicing on legal challenge platforms
- Writing up solved challenges
Not for
- Any testing of systems you do not own or have permission to test
- General application development
What's inside depth 4/5
12 skills averaging 17 KB with roughly 200 reference files, a handful of scripts and 46 eval files.
- Skills
- 12
- Reference docs
- 124
- Scripts
- 8
- Tests / evals
- 46
- Templates
- 0
- Avg SKILL.md
- 17.1 KB
Counted from the repository file tree. How we score depth.
Dependencies and costs
- Python packages and system tools listed per skill (for example pwntools, torch for ctf-ai-ml)
- TAVILY_API_KEY for ctf-similar-search
- Bash, internet access for tool installation
- Telemetry: none found; ctf-similar-search sends your search queries to Tavily when you provide a key.
Security review
Broad accessThis skill asks for more access than most, or sends data to another service. That is usually a design choice rather than bad intent, but you should know before you install it.
- Pre-approves broad shell commands13 skills set allowed-tools to plain
Bashtogether with Read, Write, Edit, Task, WebFetch and WebSearch (unrestricted shell and web access); examples are ctf-web, ctf-pwn, ctf-malware and solve-challenge.ctf-web/SKILL.md:6
A set of CTF (capture-the-flag) skills covering web, pwn, crypto, forensics, malware analysis and OSINT, written for authorised competitions and research. Thirteen skills pre-approve unrestricted Bash plus web fetch and search, and an optional installer adds many security tools using sudo. Using the skills means the agent runs offensive-security tools against targets you name, so point them only at systems you are authorised to test.
Worth knowing
- Installs packages or toolsAn optional 40 KB installer pip-installs many security tools (mostly with version pins), apt/brew packages and Go tools, uses sudo for apt, and clones some tools into ~/.ctf-tools; backends can be skipped.scripts/install_ctf_tools.sh:472
- Relies on a CLI or package we could not reviewThe skills rely on third-party security tooling (pwntools, sqlmap, nmap, hashcat, Ghidra, radare2 and similar) that is not in the repository.scripts/install_ctf_tools.sh:232
- Makes network requestsThe skills are offensive-security CTF guides; their examples send requests to target hosts the user names (placeholder domains such as target.com), and some helper scripts fuzz or probe those targets.ctf-web/SKILL.md:6
What we checked
The repository (286 files) on Oct 29, 2026. Not a full review: we read 1 of 33 code files; it relies on code outside the repository: third-party security tools installed by scripts/install_ctf_tools.sh (pip, apt, brew, Go and git clones); promptfoo and @anthropic-ai/claude-agent-sdk used by the evals; only evals/package.json and the install-command lines of install_ctf_tools.sh were reviewed; the other 32 code files (CTF helper scripts, fixtures, tests, auditor) were not read in full.
NVIDIA SkillSpector static scan: 575 pattern hits (Privilege Escalation 155, YARA Match 95, Data Exfiltration 86, Rogue Agent 65, Tool Misuse 54, MCP Rug Pull 22). The dossier hits are domain and keyword matches from CTF teaching material: placeholder attacker/target hostnames, injection and credential-handling topics, a regex list in the repository's own skill_security_auditor.py that detects curl | sh, and an Anthropic key used only by an optional maintainer CI workflow.
A review, not a certification. Skills change after we check them — read the files before installing.
License
MIT — commercial use allowed.
Can it run unattended?
Not really. Challenge solving is exploratory and the skills include pivot guidance, but there is no loop, stop condition or verification harness.
Alternatives to ctf-skills
FAQ
What is ctf-skills?
It is a set of 12 agent skills holding techniques for capture-the-flag challenges across web, binary exploitation, crypto, reverse engineering, forensics, OSINT, malware and AI/ML categories.
How do I install ctf-skills?
Run npx skills add ljagiello/ctf-skills. Each skill lists its own prerequisites, and the README also describes a central tool installer for pre-competition setup.
Is ctf-skills safe to use?
The files are technique documentation and small helper scripts, and we found no hidden calls. The content is for competitions and authorized testing, so run tools and payloads only against systems you have permission to attack.
Sources: GitHub repository (stars, license, file tree, last push Oct 27, 2026). Review based on the README and SKILL.md files as published.