Software architect, 19 years. NetSuite 10+ (SuiteScript, SuiteQL, SCA, integrations), .NET 9 years. Security engineer shipping Android apps at MetaCyberGuru LLC. I build agent skills for Android launches, app promo videos and software estimation.
by Muhammad Azhar
Your agent splits one feature into 3 to 8 tasks, gives each an optimistic, likely and pessimistic hour estimate, and returns a quote range with assumptions a client can read. A small script does the arithmetic, so every total adds up.
by Muhammad Azhar
Your agent runs one offline Python script on an AI-written diff and gets a counted, line-numbered list of deleted tests, dropped assertions, new skip markers, suppressions, CI edits that tolerate failure and files outside the agreed scope. It then triages each finding with you and reruns.
by Muhammad Azhar
Your agent runs one offline Python script on a HAR, JSON capture or log and writes a copy with tokens, cookies, JWTs, API keys, passwords, emails, IPs and card numbers swapped for consistent placeholders. A HAR stays valid JSON. You get a summary by category and a verify pass before you send it.
by Muhammad Azhar
Your agent runs one offline Python script over CLAUDE.md, AGENTS.md, GEMINI.md, Cursor and Copilot rule files. You get an estimated token range per session, a map of duplicated blocks, dead references, machine-specific paths and contradiction candidates, then merges you approve one at a time.
by Muhammad Azhar
Your agent runs one offline Python check on a SKILL.md package and your listing draft, then gives you a scorecard and an ordered fix list: case-mismatched paths, missing and orphan files, shell-only helpers, undeclared script behaviour, context weight, and listing bullets with nothing behind them.
by Muhammad Azhar
Before a repo, template or skill goes public, your agent scans it offline for client names you list, internal hosts, home paths with a username, ticket ids, document authors, photo GPS and commit identities, then writes a scrubbed copy with consistent pseudonyms and verifies it.
by Muhammad Azhar
Your agent maps a .NET Framework solution before migration: each project typed and tiered, blockers with file and line, a dependency-ordered move sequence, and web.config ported to an appsettings.json skeleton with a section mapping table. Offline and read-only. Needs only Python.
by Muhammad Azhar
Your agent scans an Android project and drafts each Play Data safety row with the dependency, permission or source line behind it, then diffs the draft against your current declaration. SDKs outside its dated map are listed as unknown. You review and submit the form yourself. Needs only Python.
by Muhammad Azhar
Your agent decodes obfuscated Android stack traces offline with the build's mapping.txt: original class, method, file and line per frame, inlined frames expanded, and a warning when the mapping is from another build. It ranks crashes by count and audits classes loaded by name. Needs only Python.
by Muhammad Azhar
Your agent checks SuiteScript 2.x against a bundled catalog of real N/ modules, members, enum values and entry points before upload. Each invented or misplaced name gets its line and the closest real name. Names the catalog doesn't cover are marked unverified, not wrong. Offline, Python only.
by Muhammad Azhar
Your agent runs a static analyzer on SuiteScript 2.x and reports, for each entry point, the fixed units, units per loop iteration, the script type limit and how many lines fit under it, with line numbers and a named fix for each hot spot. An estimate from an editable cost table. Runs offline.
by Muhammad Azhar
Your agent runs a structured OSINT campaign: define objectives and OPSEC, enumerate domains, IPs and people, pivot through passive DNS, Certificate Transparency, Shodan and archives, then correlate and preserve evidence with SHA-256 hashes and UTC timestamps. It keeps a clean chain of custody.
by Muhammad Azhar
Your agent runs a speed-ordered web, API and cloud checklist: surface mapping, auth and session flows, access control (IDOR, BOLA, BFLA), and the high-impact input bugs (SQLi, XSS, SSRF, SSTI, upload chains). It keeps a golden smoke-test set and flags the quick wins first.
by Muhammad Azhar
Your agent turns raw engagement notes into a client-ready report: a risk-led executive summary, findings with reproduction steps and impact, CVSS v3.1 and v4.0 with justified vectors, and clean evidence. It separates client and customer deliverables, redacts credentials, and tracks retest status.
by Muhammad Azhar
Your agent maps a target's attack surface, hunts dangerous functions and traces taint with CodeQL, Semgrep and Weggli, and follows untrusted input to the sink. It builds focused fuzzing harnesses and patch-diffs N-days, and reports each finding with the data-flow path, not just a grep hit.
by Muhammad Azhar
Your agent classifies a memory or logic bug, traces the object lifecycle and trust boundary, and names the primitive it yields, grounded in real CVE case studies across six core classes. It scores impact with CVSS v4.0 and never claims a primitive the code path can't support.
by Muhammad Azhar
Your agent fingerprints a target's protections (ASLR, DEP/NX, RELRO, stack canaries, CFI, seccomp), says which layer blocks your primitive, and names the prerequisite you need to get past it. It writes the defensive hardening side too, and keeps bypass work in an isolated lab.
by Muhammad Azhar
Your agent reads an ASAN, UBSAN, WinDbg or GDB crash, separates the detection point from the root cause, and gives an exploitability verdict with the controlled registers and the write it yields. It deduplicates a crash corpus and labels anything it could not reproduce.
by Muhammad Azhar
Your agent audits a Windows target's mitigations (DEP, ASLR, CFG, CET shadow stack, SEHOP, ACG, Heap Guard), maps each crash exception code to the mitigation that fired, and shows which classic primitive it kills. It builds the per-binary and system hardening picture before any bypass is designed.
by Muhammad Azhar
Your agent sizes up every request first. Small jobs run directly; big ones are split into workstreams with clear file ownership, the cheapest model that can do each part, checkpoints that survive a crash, and one final verification pass. Works in Claude Code, Codex and Cursor.
by Muhammad Azhar
Turn a brief, ticket list, repo or old estimate into a task-by-task estimate that assumes AI coding agents do the repetitive work. It traces the real code path first, keeps human hours apart from agent runtime, never double-counts QA or PM, and can write straight into a Google Sheet.
by Muhammad Azhar
Your agent takes an existing Android app through Firebase, GA4, Remote Config, Crashlytics, FCM and AdMob setup, a release build, a git tag, the Play upload and its policy declarations, then checks what Play actually shows. Nothing goes live without your yes.