Cybersecurity RL Environment Reviewer — Offensive & AppSec

Part-Time, Remote

About idlexperts

idlexperts is idler's quality assurance network — the part-time, remote contributors ("experts") who review idler's reinforcement-learning environments before they're delivered to customers. Experts are the last check between what engineering builds and what a customer receives: they confirm a task is solvable as given, confirm the grading actually tests the right thing, and flag anything that would make the task unfair or ambiguous to an AI model attempting it.

About the Role

We're building a dedicated bench of cybersecurity reviewers to QA our offensive-security and application-security RL environments: vulnerability-discovery tasks, vulnerability-patching tasks, and end-to-end exploit-construction tasks. You'll work through environments we've already built — not build new ones — and tell us whether they're fair, correct, and ready to ship.

This is part-time, flexible-hours part-time work. Most experts work a handful of hours per week; we'll scope specific weekly workload with you directly.

What You'll Do

• Review a codebase and a described vulnerability (e.g., SQLi, IDOR, SSRF, insecure deserialization, a memory-safety bug) and confirm it's real, reachable, and findable by a skilled engineer using only the context provided — no outside information required.

• For patching tasks, confirm the grading actually fails while the exploit still works and passes only once it's genuinely fixed — not just pattern-matching a diff.

• For exploit-construction tasks, trace or attempt the intended exploit path yourself to confirm it's achievable within the given environment and constraints.

• Flag inconsistent files, missing context, unclear requirements, or anything that would make the task unsolvable or ambiguous for a model attempting it cold.

• Write a short, clear justification for every accept/reject/flag decision — this matters as much as the decision itself.

What We're Looking For (Minimum Qualifications)

• Hands-on experience finding and validating real vulnerabilities in web applications, APIs, or backend services, through bug bounty, penetration testing, red teaming, or AppSec engineering.

• Comfortable reading unfamiliar code (any of Python, JS/TS, Java, Go, C/C++, Rust) well enough to trace a vulnerable data path, even without being a professional software engineer in that language.

• Able to clearly explain, in writing, why a vulnerability is or isn't real and reachable — you'll be judged on your justifications, not just your verdicts.

• Reliable and self-directed, comfortable working async with limited day-to-day oversight.

Nice to Have

• An active or past bug bounty track record (HackerOne, Bugcrowd, Synack, Intigriti) — leaderboard ranking, hall-of-fame mentions, or CVE credits are all useful signal.

• CTF experience (web or pwn categories), especially with a public CTFtime team ranking.

• A security certification such as OSCP, OSWE, OSCE, eJPT, eCPPT, or GWAPT.

• Prior experience specifically evaluating or grading other people's security findings (triage, review, or mentorship work).

Compensation

Hourly, part-time. Pay depends on experience and track record:

• $65–$100/hour — solid, demonstrable offensive-security experience (consistent bug bounty activity, 1+ years in an AppSec/pentest role, or a relevant certification) and the ability to independently verify straightforward discovery and patching tasks.

• $125–$250/hour — a recognized track record (top-tier bug bounty leaderboard placement, live-hacking-event invitations, advanced certifications such as OSWE/OSCE, or 5+ years of specialized offensive-security work), able to handle ambiguous or disputed calls and the hardest exploit-chaining tasks.

Exceptional, elite-tier reviewers (e.g., top-50 global bug bounty hunters) may be engaged above this range case-by-case for our highest-difficulty environments.

How to Apply

Share links to your bug bounty profile(s), CTF team page, GitHub, or any public write-ups/CVEs. We'll follow up with a short, paid sample review (one task, roughly 30–45 minutes) as the final step before starting regular work.