Agent skillsVertical agents

reverse-skill

zhaoxuya520/reverse-skill

A routing layer that tells your coding agent which security tool to reach for, and how to use it properly.

Repository
Stars
22.8k
Forks
2.1k
Open issues
34
Language
PowerShell
Licence
MIT
Created
2026-05-13
Last push
2026-08-09
Effort to adopt

Some wiring

A few days of integration work — credentials, data mapping, a deploy target.

01

What it actually does

Coding agents are confident and frequently wrong about specialist tooling. Hand one an Android APK, a stripped binary, an obfuscated JavaScript bundle, or a packet capture and it will guess at commands — reaching for jadx when it needed Frida, or inventing flags that do not exist.

reverse-skill sits in front of that guessing. It is a router: 41 rules that read the task, decide which methodology applies, check whether the required tools are actually installed on the machine, bootstrap anything missing, and then run a repeatable playbook instead of improvising. It ships 42 tracked skill modules and a 163-case regression benchmark, and it runs on Windows and Ubuntu CI.

The second half is the part teams underestimate — a self-evolving knowledge base. Findings, evidence, and dead ends get written back into a field journal, so the same mistake is not repeated on the next engagement. It works with Claude Code, Kiro, Cursor, Cline, and other agent clients.

Important framing: this is built for authorized work. The workflow forces a scope file with explicit authorization and a network profile before any action is taken against a target.

02

Who it's for

  • 01

    Security teams running authorized penetration tests who want consistent methodology instead of whatever the analyst remembered that day

  • 02

    Application security engineers at SaaS companies reviewing their own mobile and web builds before release

  • 03

    Engineering leaders who need security findings documented in a repeatable, auditable format

  • 04

    CTF players and security researchers who want tool selection to stop being the bottleneck

EngineeringOperationsFoundersSaaSFintechAny industry
03

Where it earns its keep

  • Triaging your own mobile app build to see what an attacker would extract from it
  • Standardising how a security team documents evidence, findings, and reproduction paths
  • Onboarding a junior analyst who knows the theory but not which of forty tools to open first
  • Running a repeatable pre-release check against a binary or bundle as part of CI
  • Building an institutional memory of what worked, so engagement two is faster than engagement one
04

Use it, or skip it

Reach for it when

  • You have written authorization to test the target, and scope is defined
  • Your team keeps re-deriving the same methodology because nothing is written down
  • Tool sprawl is real — scripts and MCP servers scattered across different machines
  • You want an agent to follow a playbook, not free-associate about security

Skip it when

  • You do not have explicit written authorization for the target. The tooling assumes you do; that is your responsibility, not the router's
  • You need a compliance-grade audit with a signed report — this accelerates practitioners, it does not replace a qualified assessor
  • Your team has no security background at all. This routes expertise, it does not create it, and misread findings are worse than none
  • You want an off-the-shelf SaaS scanner with a dashboard — this is a practitioner's toolkit that lives in the terminal
05

10 automations

  1. 01Engineering

    Pre-release mobile build check

    On every release-candidate build, run the APK playbook against your own artifact and post a findings summary into the release channel before the ship decision is made.

  2. 02Engineering

    Evidence-to-ticket pipeline

    Convert each Evidence→Finding→Path record into a Jira or Linear ticket with severity, reproduction steps, and the affected file already filled in.

  3. 03Founders

    Security posture digest for the founder

    Weekly plain-English summary of what was tested, what was found, and what is still open — no jargon, delivered to Slack every Friday.

  4. 04Engineering

    Dependency exposure sweep

    When a new CVE lands for a library you ship, route an automated check across your own builds and report which artifacts actually contain the vulnerable path.

  5. 05Operations

    Scope guard enforcement

    Block any agent action until a signed scope file with authorization and network profile exists, and log every attempt that was refused.

  6. 06Operations

    Field journal to knowledge base

    Sync the self-evolving journal into your internal wiki nightly so the security playbook becomes searchable by the whole engineering org.

  7. 07Operations

    Onboarding path for new analysts

    Generate a personalised ramp plan from the 42 skill modules based on which ones a new hire has not yet used, tracked as a checklist.

  8. 08Engineering

    Toolchain drift alerts

    Nightly check that every machine in the team has the same tool versions bootstrapped, and open a ticket when one drifts.

  9. 09Sales

    Customer security questionnaire assist

    Pull real evidence from past authorized tests to pre-fill the security sections of enterprise procurement questionnaires, with a human approving each answer.

  10. 10Engineering

    Regression benchmark gate

    Run the 163-case benchmark on any change to your routing rules and refuse to merge when the pass rate drops.

Want one of these running by Friday?

LimeDock builds these as real workflows inside your stack — deployed to your cloud, wired into your Slack and CRM, with the code in your repo. You pay a build fee and your own API keys, nothing else.

Book a workflow call
07

Source

Repository stats were read from the GitHub API and reflect the last time we refreshed this entry. The editorial breakdown above is LimeDock’s own analysis — we are not affiliated with zhaoxuya520.

https://github.com/zhaoxuya520/reverse-skill