Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/Encod3d-Sec/TORCHnpx agentmods add skills/encod3d-sec/torch/ctf-workflowWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/encod3d-sec/torch/ctf-workflow)<a href="https://agentmods.dev/skills/encod3d-sec/torch/ctf-workflow"><img src="https://agentmods.dev/badge/skills/encod3d-sec/torch/ctf-workflow/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/encod3d-sec/torch/ctf-workflow"><img src="https://agentmods.dev/badge/skills/encod3d-sec/torch/ctf-workflow.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 3 findings, up to critical
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- critical YARA Match · line 35 YARA rule matched a known malware signature (reverse shell, backdoor, ransomware, C2 framework, or info stealer).Fix: Remove the malware payload or compromised file entirely. Investigate how it entered the skill and audit all other artifacts for additional indicators of compromise.
- high YARA Match · line 31 YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).Fix: Remove offensive tool references and exploit code. Legitimate agent skills should not contain penetration testing tools, exploit frameworks, or reconnaissance utilities.
- medium Excessive Agency · line 94 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00124 | $0.01950 |
| Opus 5.5 | $0.00050 | $0.00780 |
| Sonnet 5.5 | $0.00025 | $0.00390 |
| Haiku 4.5 | $0.00012 | $0.00195 |
Grade A, and why
ctf-workflow scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
its DOM + network requests (the rendered XHR/fetch calls reveal API routes a `curl` crawl misses - How it starts
The opening of the file, as written. The whole thing — 115 lines — stays where its author put it; the contents beside it link to each section on GitHub.
ctf-workflow
The driver is the plan. Run one command, do exactly what it prints, record the result, repeat. The
gates are enforced by scripts/campaign.py.
The loop
python3 scripts/campaign.py next
python3 scripts/campaign.py note <row> --arsenal <slug>
python3 scripts/campaign.py done <row> --poc <img> --kind req # | --dead R | --park Q | --find F
Every next now prints an APPROACH:/AVOID:/REFS: block for the served row (the distilled
ctf-box method for that vuln class) - read it before acting.
Start / resume
python3 scripts/campaign.py init --type ctf- validatesscope.md(for a box, scope is just the target IP/host) + envelope.- OSINT is OFF by default for CTF (pass 0 is skipped). Only run it if this skill was invoked
with an explicit
osintargument - a box's answer is on the box, not in Wayback. - Passes 1-3 feed
state.md- rustscan/nmap + web enum. Read every service banner, page source and config end to end. python3 scripts/campaign.py board- writes the 4a foothold rows; once an asset is a foothold, re-runboardand it seeds the 4b privesc rows (pspy/linpeas auto + the manual checklist) for that asset.- Enter the loop, depth-first.
- When a foothold lands (a reverse shell in a tmux window via
vm-scan.sh --win shell, or a meterpreter/msfconsole session via--win msf), record it:python3 scripts/campaign.py foothold <target> --win shell(or--win msf; or ride it on the closing find withdone ... --win). The driver flips the asset'sstate.mdrow toaccess=footholdand routes its 4b privesc rows throughvm-rsh --win <win>(persistent session + operator visibility past foothold), and printstmux attach -t <eng>for manual takeover. msf itself is operator-attach / drop-to-shell, notvm-rsh-driven (its wrapper frames a bash shell, not themsf6 >REPL). After recording a foothold, re-runpython3 scripts/campaign.py boardso the 4b privesc rows are seeded -nextwill not surface them until you do. - Web RCE -> a real shell, THEN stabilize -- do not ride one-liners (recurring drift). The
moment code-exec lands (a web-RCE primitive, an LFI->session-poison, a deser gadget), STOP
hand-poking one-shot payloads: (a) catch it with a real handler by default -
msfconsole'sexploit/multi/handler(meterpreter first; a plainshell_reverse_tcpis the backup when meterpreter is blocked, e.g. Windows/EDR) (record viacampaign.py foothold <target> --win msf, step 6). Reserve a rawnc -lvnplistener for when msf is unavailable: a raw nc pane'sCtrl-Ckills the LISTENER (dropping the shell back to your own prompt, the false-root attacker-prompt trap) and it has no session management, while meterpreter also carriespost/multi/recon/local_exploit_suggesterescalation modules and built-in file transfer. (b) Before picking the LPORT, test target egress on common ports (80/443/53) - high ports like 4444 are frequently filtered, so pick an egress-allowed LPORT. (scripts/vm-handler.sh <eng> <lhost>picks a free egress port and launches the handler for you, printing the LPORT.) (c) If you did fall back to raw nc, stabilize it immediately withbash scripts/vm-stabilize.sh --win shell <eng>(pty + job control + window size). (d) Then record the foothold (step 6) and drive withvm-rsh. An unstabilized nc shell (no job control, mid-line wrapping) is what makes post-ex drift back into one-liners. Full discipline:Skill(ctf-box)Phase 3 (Deliver). Therecon-capturehook fires this reflex once on the first service-accountid.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago First seen · 115 lines · 124 tokens per session scan A 4a7c6817ad44
ctf-workflow is a skill published in the GitHub repository Encod3d-Sec/TORCH (336 stars, last pushed 1mo ago), licensed MIT. It adds 124 tokens to every session and 1,950 once invoked, about $0.0005 per session on Opus 5.5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
local-ai-agents
Build local-first AI agents that run entirely on a developer workstation with Microsoft Foundry Local and Qwen function-calling models. Covers Small Language Models (SLMs), the OpenAI-compatible local endpoint, sandboxed local tools, local RAG with Chroma, local MCP servers, hybrid cloud/local routing, and the…
systematic-debugging
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.
next-cache-components-adoption
Turn on Cache Components in a Next.js app and resolve the blocking routes it surfaces. Use when the user wants to enable, adopt, or migrate to Cache Components, flip the cacheComponents flag, work through a flood of blocking-prerender / instant validation errors, run the cache-components-instant-false codemod, or…
chat-pet-sprite-creation
Use when creating or changing VS Code chat pet sprite art, sprite sheets, state animations, eye treatments, Stable/Insiders variants, or pet transitions under src/vs/workbench/contrib/chat/browser/widget/media/chatPet.
cpu-profile-analysis
Analyze V8/Chrome CPU profiles (.cpuprofile) and DevTools trace files (Trace-.json). Use when: profiling performance, investigating slow functions, comparing code paths, finding bottlenecks, analyzing timeToRequest, understanding call trees from sampling profiler data, analyzing layout/paint/rendering, investigating…
insight-error-page
Write or audit an insight-kind error page for the Next.js dev overlay. Use when creating a new errors/ .mdx page, auditing an existing one, or checking that a page matches the framework fix cards. Covers page structure, title alignment, FixCard cards with Copy prompt button, code snippets, terminology verification…