How do I install the Hunt Deserialization skill?

Run npx skillmds@latest add duriantaco/hunt-deserialization in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Hunt Deserialization safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Hunt Deserialization?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Hunt Deserialization free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Hunt Deserialization?

duriantaco (@duriantaco) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Hunt Deserialization by duriantaco — Agent Skill

Name: Hunt Deserialization
Author: duriantaco

Hunt Deserialization

Investigate authorized unsafe deserialization and serialized-data trust boundaries. Use when typed discovery identifies encoded object cookies or tokens, pickle or object streams, YAML type tags, serialized uploads or imports, type-confusion errors, or suspected server-side object reconstruction.

duriantaco 19ae455 2 files · 1.8 KB Updated Sep 17, 2026

File contents

Hunt Deserialization Boundaries

Treat this card as advisory prioritization. Encodings, format fingerprints, and parser errors do not prove unsafe object construction.

Workflow

Identify the format, carrier, integrity protection, identity, and target-observed replay.
Pair an unchanged baseline with a same-shape inert structural control; change one property.
Use cookie_deserialization for cookies or tokens and file_fetch_parser only for typed imports or uploads. Their fixed command and file-read candidates must all be authorized.
Preserve redacted inputs, format decisions, controls, deltas, and evidence references. Report a validated vulnerability independently of any challenge objective.

Evidence Gate

Confirm only a replayable unsafe type, object-handling, or execution effect absent from the paired control. Decoder errors, library names, gadget strings, and status changes remain observations. Keep any contract_missing result suspected rather than promoting it.

Stop Conditions

Stop when format, integrity, identity, replay, or controls are unavailable, or policy limits fire. Do not add destructive gadgets, persistence, or outbound effects; redact returned secrets.

Frequently asked questions