reverse-engineer-js
Reverse-engineer minified or obfuscated JavaScript bundles. Primary tool: humanify (humanifyjs) — uses an LLM to rename mangled identifiers back to meaningful names. Default model: gemini-3.1-flash-lite-preview. Use when the user says 'deobfuscate this bundle', 'humanify this', 'rename minified vars', 'reverse engineer this JS', 'what does bundle.js do', or is working in a project with a deobfuscated `og/` or `modules/` folder fed by humanify. Covers humanify CLI invocation, the pseudo-TTY workaround for cursorTo crashes, single-module extraction from a bundle, and chaining with beautifiers.
What this skill does
# Reverse Engineer JS
LLM-assisted deobfuscation of minified / mangled JavaScript using [humanifyjs](https://github.com/jehna/humanify) plus surrounding tooling for extracting and reintegrating modules.
# Tools at a glance
| Tool | Purpose | When |
|------|---------|------|
| `humanifyjs` (npx) | Rename mangled vars/functions via LLM | Mangled but already valid JS |
| `prettier` / `js-beautify` | Reformat single-line minified code | Before humanify, if needed |
| `webcrack` | Unpack webpack / browserify bundles into per-module files | When bundle is one big file |
| Source-map sleuthing | `// # sourceMappingURL=...` may exist next to the bundle | Always check first — beats LLM rename |
Always check for a sourcemap (`bundle.js.map` or `sourceMappingURL` comment in the file) **before** spending tokens on humanify. If found, original sources are recoverable for free.
# Default humanify invocation
Use Gemini with the lite preview model unless the user says otherwise:
```bash
npx humanifyjs gemini -m "gemini-3.1-flash-lite-preview" -o <outputDir> <inputFile>
```
Requires `GEMINI_API_KEY` in env (the project's root `.env` is the usual place).
Other backends:
```bash
# OpenAI
npx humanifyjs openai --apiKey="$OPENAI_API_KEY" --model gpt-4.1-mini bundle.js
# Local (slower, free)
npx humanifyjs local bundle.js
```
# Critical gotchas
## 1. Pseudo-TTY required in non-interactive shells
humanify calls `process.stdout.cursorTo(...)` for its progress bar. In Claude Code / CI / piped contexts, stdout is not a TTY and the call **crashes with `TypeError: process.stdout.cursorTo is not a function`**.
Wrap the call in `script -q /dev/null`:
```bash
script -q /dev/null npx humanifyjs gemini -m "gemini-3.1-flash-lite-preview" -o "$OUT" "$IN"
```
(macOS/BSD `script` syntax. On Linux: `script -qc "<cmd>" /dev/null`.)
## 2. Output filename is fixed
humanify always writes its result as `<outputDir>/deobfuscated.js` regardless of input filename. Don't expect `module_1234.js` in → `module_1234.js` out.
## 3. Long files = long runs
humanify processes top-level declarations sequentially with one LLM call each. A 5 MB bundle can take 30+ minutes. Strategies:
- Run on **single modules** extracted from the bundle, not the whole thing
- Pre-split with `webcrack` first if the bundle is webpacked
- Use the cheapest competent model (`gemini-3.1-flash-lite-preview` is the current sweet spot)
## 4. Keep an archive of the original
Before any rewrite, copy the pristine bundle to `archive/bundle.original.js`. humanify is non-deterministic and you may want to re-run with a different model.
# Workflow A: Whole-bundle deobfuscation
When the bundle is small (<500 KB) or you have time:
```bash
mkdir -p archive out
cp bundle.js archive/bundle.original.js
script -q /dev/null npx humanifyjs gemini \
-m "gemini-3.1-flash-lite-preview" \
-o out bundle.js
# result: out/deobfuscated.js
```
Then optionally run `prettier --write out/deobfuscated.js` for nicer formatting.
# Workflow B: Single-module extraction (large bundles)
Common in fxhash / generative-art bundles: a single bundle contains many `id: (params) => { ... }` modules and you only want to humanify a few interesting ones.
1. **Identify the module ID** — search the bundle for distinctive strings, then trace back to the enclosing `<id>: (e, t, n) => {` opener.
2. **Extract the module body** — match the opener, count braces (respecting strings) until the matching `}`.
3. **Wrap as a standalone arrow** — write `(e, t, n) => { <body> }` to a temp file. humanify needs valid top-level code.
4. **Run humanify** — `script -q /dev/null npx humanifyjs gemini -m "gemini-3.1-flash-lite-preview" -o "$OUT" "$IN"`.
5. **Unwrap** — strip the `(e, t, n) => { ... }` shell back off the result.
6. **Save and stub** — write the deobf'd body to `modules/module_<id>_<name>.js`, replace the body in the bundle with a stub like `<id>: (e, t, n) => { window.DEOBF[<id>](e, t, n); }`.
If the project already has a `scripts/deobfuscate.js` (or similar), reuse it — the brace-counting / string-escape logic is finicky enough that re-deriving it for each new bundle is a waste. A typical script: takes a numeric module ID, extracts the body from `bundle.js`, runs humanify on the wrapped function, unwraps, and writes `modules/module_<id>_<name>.js`.
# Workflow C: Webpack bundle → per-module files
When dealing with a webpack bundle, [webcrack](https://github.com/j4k0xb/webcrack) splits modules cleanly before humanify even runs:
```bash
npx webcrack bundle.js -o webcracked/
# Then humanify each module of interest:
script -q /dev/null npx humanifyjs gemini \
-m "gemini-3.1-flash-lite-preview" \
-o humanified webcracked/modules/123.js
```
# Examples
## Example 1: deobfuscate a small fxhash bundle
**User says:** "humanify the bundle.js in this fxhash project"
**Actions:**
1. Check for sourcemap: `head -c 4096 bundle.js | tail -c 200` — look for `sourceMappingURL`. None? Continue.
2. Confirm `GEMINI_API_KEY` in `.env`. Source it: `set -a; source .env; set +a`.
3. Archive: `mkdir -p archive && cp bundle.js archive/bundle.original.js`.
4. Run: `mkdir -p out && script -q /dev/null npx humanifyjs gemini -m "gemini-3.1-flash-lite-preview" -o out bundle.js`.
5. Beautify: `npx prettier --write out/deobfuscated.js`.
**Result:** `out/deobfuscated.js` with renamed vars, ready to read.
## Example 2: deobfuscate one module out of a giant bundle
**User says:** "Deobfuscate module 3658 from `og/bundle.js`, name it `treeGenerator`"
**Actions:**
1. If the project already has a deobfuscation helper (e.g. `scripts/deobfuscate.js`), run it: `node scripts/deobfuscate.js 3658 treeGenerator`.
2. Otherwise: extract the module body by brace-counting, wrap as `(e, t, n) => { ... }`, run humanify with `script -q /dev/null`, unwrap, save to `modules/module_3658_treeGenerator.js`, stub the bundle.
**Result:** New file in `modules/` with a readable version of the function; `bundle.js` shrunk by stubbing.
## Example 3: humanify crashed with `cursorTo is not a function`
**User says:** "humanify just dies immediately"
**Diagnosis:** Non-TTY stdout (this is the default in Claude Code's bash tool).
**Fix:** Re-run wrapped in `script -q /dev/null <cmd>` (macOS) or `script -qc "<cmd>" /dev/null` (Linux).
# Troubleshooting
## `cursorTo is not a function`
**Cause:** stdout is not a TTY. **Fix:** wrap in `script -q /dev/null`.
## `GEMINI_API_KEY` not set
**Cause:** humanify reads from process env, not from the project `.env` automatically. **Fix:** `set -a; source .env; set +a` in the same shell, or pass inline: `GEMINI_API_KEY=... npx humanifyjs gemini ...`.
## Output is identical to input (no renames)
**Cause:** humanify only renames identifiers it considers mangled (short, single-letter). Already-readable code is left alone. **Fix:** confirm the input is actually obfuscated; if names like `useState` survive that's expected, not a bug.
## Rate limits / quota errors from Gemini
**Cause:** large file, many top-level functions → many sequential LLM calls. **Fix:** switch to `local` backend for free local rename, or extract a single module (Workflow B) instead of the whole bundle.
## humanify takes hours
**Cause:** big bundle, sequential per-function calls. **Fix:** pre-split with `webcrack`, or only humanify the modules you actually need.
## Output directory contains stale `deobfuscated.js`
**Cause:** humanify writes to a fixed filename. A failed second run will pick up the previous good output. **Fix:** `rm -f $OUT/deobfuscated.js` before each run.
Related in AI Agents
skill-development
IncludedComprehensive meta-skill for creating, managing, validating, auditing, and distributing Claude Code skills and slash commands (unified in v2.1.3+). Provides skill templates, creation workflows, validation patterns, audit checklists, naming conventions, YAML frontmatter guidance, progressive disclosure examples, and best practices lookup. Use when creating new skills, validating existing skills, auditing skill quality, understanding skill architecture, needing skill templates, learning about YAML frontmatter requirements, progressive disclosure patterns, tool restrictions (allowed-tools), skill composition, skill naming conventions, troubleshooting skill activation issues, creating custom slash commands, configuring command frontmatter, using command arguments ($ARGUMENTS, $1, $2), bash execution in commands, file references in commands, command namespacing, plugin commands, MCP slash commands, Skill tool configuration, or deciding between skills vs slash commands. Delegates to docs-management skill for official documentation.
reprompter
IncludedTransform messy prompts into well-structured, effective prompts — single or multi-agent. Use when: "reprompt", "reprompt this", "clean up this prompt", "structure my prompt", rough text needing XML tags and best practices, "reprompter teams", "repromptception", "run with quality", "smart run", "smart agents", multi-agent tasks, audits, parallel work, anything going to agent teams. Don't use when: simple Q&A, pure chat, immediate execution-only tasks. See "Don't Use When" section for details. Outputs: Structured XML/Markdown prompt, quality score (before/after), optional team brief + per-agent sub-prompts, agent team output files. Success criteria: Single mode quality score ≥ 7/10; Repromptception per-agent prompt quality score 8+/10; all required sections present, actionable and specific.
adaptive-compaction
IncludedAdaptive add-on policy and recovery layer that decides WHEN to compact, prune, snapshot, or fork -- replacing fixed-percent auto-compaction across Claude Code, Codex, and MCP-capable hosts. Trigger on auto-compact timing or damage: "when should I compact", "is it safe to compact now or start a fresh session", "auto-compact fires too early/mid-task", "switching to an unrelated task but the window still has space", "context rot", "answers get worse the longer the session runs", "the agent forgot the plan or my decisions after it summarized", "add a layer on top that manages context without changing the agent", raising autoCompactWindow to give the policy room, or installing/tuning a cross-tool compaction policy or PreCompact hook -- even when "compaction" is never said but the problem is context-window pressure or post-summarization memory loss. Do NOT use to summarize a conversation, build RAG, write a summarization prompt (decides WHEN not HOW), or answer max-context-length trivia.
agent-skill-creator
IncludedCreate cross-platform agent skills from workflow descriptions. Activates when users ask to create an agent, automate a repetitive workflow, create a custom skill, or need advanced agent creation. Triggers on phrases like create agent for, automate workflow, create skill for, every day I have to, daily I need to, turn process into agent, need to automate, create a cross-platform skill, validate this skill, export this skill, migrate this skill. Supports single skills, multi-agent suites, transcript processing, template-based creation, interactive configuration, cross-platform export, and spec validation.
llm-wiki
IncludedUse when building or maintaining a persistent personal knowledge base (second brain) in Obsidian where an LLM incrementally ingests sources, updates entity/concept pages, maintains cross-references, and keeps a synthesis current. Triggers include "second brain", "Obsidian wiki", "personal knowledge management", "ingest this paper/article/book", "build a research wiki", "compound knowledge", "Memex", or whenever the user wants knowledge to accumulate across sessions instead of being re-derived by RAG on every query.
skill-master
IncludedAgent Skills authoring, evaluation, and optimization. Create, edit, validate, benchmark, and improve skills following the agentskills.io specification. Use when designing SKILL.md files, structuring skill folders (references, scripts, assets), ingesting external documentation into skills, running trigger evals, benchmarking skill quality, optimizing descriptions, or performing blind A/B comparisons. Keywords: agentskills.io, SKILL.md, skill authoring, eval, benchmark, trigger optimization.