rea-grilling
Grill the user relentlessly about a reverse engineering, lost source code recovery, binary triage, web/mobile security audit, or asset extraction plan before writing code. Use when the user asks to scope, plan, or stress-test an REA task, or uses trigger phrases like 'grill me', 'scope this', 'interview me', or 'plan recovery'.
How do I install this agent skill?
npx skills add https://github.com/smartworldarafath/rea_skill --skill rea-grillingIs this agent skill safe to install?
- Gen Agent Trust Hubpass
The skill is a structured interrogation engine for reverse engineering tasks. It is generally safe but carries a low risk of indirect prompt injection due to its design, which involves ingesting diagnostic tool outputs from potentially untrusted binary artifacts.
- Socketpass
No alerts
- Snykpass
Risk: LOW · No issues
What does this agent skill do?
REA Grilling: Interactive Target & Scope Interrogation Engine
Interview the user relentlessly until you reach a crystal-clear, shared understanding of the target, constraints, and success criteria. Map the investigation as a decision tree: every engineering decision branches into lower-level technical choices.
Work the tree in rounds. The frontier is every decision whose prerequisites are already settled: the questions you can ask now without guessing at answers you haven't heard yet. Ask the whole frontier in one round: number each question and give your recommended answer. Then wait for the user's answers before moving to the next round.
Format each round like this:
❓ **Q1** - **<question title>**: <question body with clear context and available choices>
➡️ **Recommended**: <your recommended technical answer based on initial artifact inspection>
---
❓ **Q2** - **<question title>**: <question body with clear context and available choices>
➡️ **Recommended**: <your recommended technical answer based on initial artifact inspection>
Crucial Rules:
- The "Yes" Rule: Word each question and recommendation so that if the user simply replies
"yes", it accepts all of your recommended answers for that round. - Finding Facts is Your Job: Never ask the user for facts you can determine yourself using tools (
readelf,DIE,strings,open_binary,inspect_pe_resources,grep,file). Inspect the target first, discover the compilers/architecture/frameworks, and then grill the user only on intent, fidelity, and scope trade-offs. - No Skeletons Guarantee: Once the grilling interview concludes and all frontier branches are settled, compile the final agreed specification and immediately execute the 5-Stage REA Pipeline under the Zero-Skeleton Law.
The REA Decision Tree Frontier
When interrogating a reverse engineering or recovery target, compute questions across these frontier branches:
Branch 1: Scope & Recovery Fidelity
- Full Idiomatic Rewrite vs. Byte-Exact Decompilation: Do you require 100% byte-exact binary matching, or clean, modern, idiomatic source code (e.g., C++20 / Rust / Go) that compiles cleanly and reproduces the exact feature behavior?
- Feature Prioritization: Are all UI modules, background workers, and protocols required, or is there a specific core subsystem to reconstruct first?
Branch 2: Toolchain & Compiler Targeting
- Target Compiler: Which compiler toolchain should the 1-click build script support?
- Options: Zig CC (
zig cc), MSVC (cl.exe), MinGW GCC, Clang, Cargo, Gradle, Go.
- Options: Zig CC (
- Dependency Strategy: Should third-party libraries (e.g., SQLite, zlib, OpenSSL, ImGui) be vendor-bundled, dynamically linked, or fetched via package managers (vcpkg, Conan, Cargo)?
Branch 3: Resource & Asset Extraction Depth
- Embedded Assets: Should binary resources be extracted as loose editable files (valid
.ico,.png,.wav,.glsl,.ttf) or linked directly into a compiled.rcresource file? - Database & Storage Schemas: Should SQLite/Realm schemas be emitted as raw SQL migration files (
schema.sql) or mapped to ORM / typed data models?
Branch 4: Security & Sensitive Artifact Handling
- Secret Sanitization: When analyzing decompiled code or web bundles, how should extracted API keys, tokens, or credentials be handled?
- Options: Auto-sanitize into
.env.exampletemplates, replace with mock test fixtures, or quarantine securely.
- Options: Auto-sanitize into
- Vulnerability Remediation: Should the output include direct inline code patches (remediations) alongside the vulnerability report?
Branch 5: Verification & Acceptance Gates
- Acceptance Criteria: What constitutes proof of 100% successful recovery?
- Options: 0 compilation errors and 0 warnings via
build.bat/build.sh, successful automated unit/smoke tests, dynamic GUI launch test, or diff coverage against original binary exports.
- Options: 0 compilation errors and 0 warnings via
Interview Execution Flow
- Initial Artifact Triage (Silent Fact-Finding):
- Run REA MCP or local inspection tools to determine: file format, architecture (x86/x64/ARM64), packing/obfuscation, compiler signature, and symbol status.
- Round 1 (Target Scope & Fidelity):
- Present findings from fact-finding.
- Ask Round 1 frontier questions (Scope, Language/Compiler target, Fidelity level) with your recommended answers.
- Round 2 (Toolchains, Resources & Edge Cases):
- Unblock dependent questions based on Round 1 answers.
- Address dependencies, asset extraction, database reconstruction, and secret handling.
- Round 3 (Verification & Final Handoff):
- Finalize build scripts, test harness requirements, and acceptance criteria.
- Execution Gate:
- Summarize the settled Recovery Specification.
- Hand off to
$rea-full-code-recoveryor the matching domain skill to reconstruct the complete codebase without any skeletons or stubs.
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/smartworldarafath/rea_skill/rea-grilling">View rea-grilling on skillZs</a>