- Home
- Skills
- Code Quality & Review
- Trading Code Review & Validation Planner
Trading Code Review & Validation Planner
Review MT5 and TradingView Pine source against explicit rules, document supported defects and evidence gaps, and plan prioritized demo-only validation.
$5
Works with the AI Agents you already use
Trading Code Review & Validation Planner
Example session with this skill installed
Review the following complete bounded units and claim C against the stated contract. Do not modify or execute MQL, connect accounts or place orders. Provide justified findings/no-findings, an evidence ledger and prioritized demo-only validation plan.
Original bounded review inputs
These complete MQL5-style functions are review units, not standalone EAs or installable trading programs. No event shell/order API is supplied. Native compilation/runtime is NOT_RUN unless independently observed. Their contract concerns arithmetic only, not broker suitability.
Contract shared by A and B
Inputs are finite: budget is positive account currency, lossPerLot is a positive all-in estimated loss in account currency per one lot, step and minimum are positive lots. Invalid input returns0. Candidate quantity is floored to step, never rounded/clamped upward. Return0 if no minimum-valid volume fits the budget. A final estimated-loss check rejects any computed volume above budget. Maximum lot, margin, fill/stop checks and real broker metadata are outside this bounded unit; they must be supplied/validated before execution. Do not add a trading risk percentage or change the budget.
A: units/ReviewedLots.mqh
double ReviewedLots(const double budget,const double lossPerLot,const double step,const double minimum)
{
if(!MathIsValidNumber(budget) || !MathIsValidNumber(lossPerLot) ||
!MathIsValidNumber(step) || !MathIsValidNumber(minimum) ||
budget<=0 || lossPerLot<=0 || step<=0 || minimum<=0) return 0;
double volume=MathFloor(budget/lossPerLot/step)*step;
if(!MathIsValidNumber(volume) || volume<minimum || volume*lossPerLot>budget) return 0;
return volume;
}
B: units/CandidateLots.mqh
double CandidateLots(const double budget,const double lossPerLot,const double step,const double minimum)
{
if(!MathIsValidNumber(budget) || !MathIsValidNumber(lossPerLot) ||
!MathIsValidNumber(step) || !MathIsValidNumber(minimum) ||
budget<=0 || lossPerLot<=0 || step<=0 || minimum<=0) return 0;
double volume=MathFloor(budget/lossPerLot/step)*step;
return MathMax(minimum,volume);
}
C: incomplete candidate/evidence
Only the text 'MT5 compile:0 errors,0 warnings' is supplied, without the EA source/includes, candidate hash, actual compiler log/build, specification, inputs, broker symbol/model/history or runtime receipts. The requester asks whether the EA is safe/live-ready. No order/chart/terminal action is authorized. Assess what the supplied evidence supports and produce a concrete handoff; do not treat this text as a verified compiler result.
- Read your context and instructions
- Compiled the trading code review
Worked source review and demo-only validation plan
Request and scope
Review A/B from review-inputs.md against their shared arithmetic contract, preserve the budget/rules, and assess C's claimed readiness. Do not execute MQL, attach charts, connect accounts or fix source. These are complete bounded function units, not whole EAs. Reviewer here is the example author: self-review, not independent certification. This worked example supplies no native compile/runtime evidence.
The exact code blocks identify units/ReviewedLots.mqh and units/CandidateLots.mqh. Line numbers count only the fenced source, starting with the function declaration at1. Obtain exact hashes from supplied bytes when doing a real review; a hash is identity, not proof of correctness. This example does not fabricate a customer candidate hash.
Disposition
| Candidate | Scoped result | Meaning |
|---|---|---|
| A | NO_SUPPORTED_FINDINGS_IN_REVIEWED_UNIT | Source reasoning and stated numerical cases support the bounded arithmetic contract. Not whole-EA, native, historical or live acceptance. |
| B | FAIL_SOURCE_CONTRACT | Demonstrated clamp-up violation exceeds the approved estimated-loss budget; block acceptance of this sizing path. |
| C | INCONCLUSIVE | A text-only compile claim cannot identify or validate the absent EA. Native compile NOT_VERIFIED, runtime NOT_RUN and live use not authorized. |
Finding R1 — HIGH
Location: units/CandidateLots.mqh:7, return MathMax(minimum,volume).
Contract: quantities are floored to step and vetoed when the minimum cannot fit; never round/clamp upward. Trigger: budget20USD, lossPerLot250USD/lot, step0.01lot, minimum0.10lot. Floor yields0.08lot; the return raises it to0.10lot. Estimated loss becomes25USD,5USD above budget (25% relative overrun). Expected quantity0/veto, actual source expression0.10. This is SOURCE_REASONING plus reproducible arithmetic, not a native MT5 observation or prediction of actual realized loss.
Minimal remedy: reject the candidate when it is below minimum; reject nonfinite computed volume and any final estimated loss above budget. Do not raise the budget, reduce an approved stop, change strategy rules or introduce risk-recovery logic. Source changes need separate authorization and a new identity/review.
Regression R1-T1: native call of the corrected unit with those four inputs must return0. R1-T2: budget25,lossPerLot100,step0.01,min0.10 must return0.25. R1-T3: budget24,lossPerLot100,step0.10,min0.10 must return0.20. Add invalid/overflow and exact budget/minimum boundary cases. All native tests remain NOT_RUN here; testing a Python mirror cannot pass them.
Evidence gaps, not additional invented defects
A's broker margin/stops/maximum volume, order ownership, protection, persistence and execution are outside the stated pure unit. Request their caller/contract before judging a complete EA; don't penalize this unit for not placing orders. C lacks complete source/includes, exact candidate identity, original compiler log/build and input/runtime context. These omissions prevent a readiness decision; they are not proof that an unseen implementation has a particular coding bug.
Prioritized handoff
| ID | Prerequisite/setup and action | Expected / pass criterion | Required artifact | Actual status |
|---|---|---|---|---|
| P1 | Preserve B and contract; reproduce20/250/0.01/0.10 arithmetic | B demonstrates0.10 vs required0/veto; estimated25>20 | Original source identity and arithmetic worksheet | SOURCE_REASONING; not native |
| P2 | Owner-authorized fix, new hash and independent source review | No supported HIGH/CRITICAL defect in corrected unit scope | Diff, source/include hashes, review dispositions | NOT_RUN |
| P3 | Installed MetaEditor5; compile a dedicated non-trading harness for exact unit | Actual compiler log; errors/warnings resolved or dispositioned | Build version, full log, exact source/binary identities | NOT_RUN |
| P4 | Separately authorized isolated Tester native harness | Native assertions for positive, minimum, equality, invalid/overflow cases pass | Native journal/assertions/settings | NOT_RUN |
| P5 | Complete actual EA/risk/execution contract and broker metadata | Whole-EA signal/risk/protection/ownership/restart checks scoped before simulation | Frozen inputs, symbol metadata, native test matrix | BLOCKED by missing full EA |
| P6 | Freeze economic plan before outcomes, complete real-tick data/costs/fills | Preregistered criteria resolved with all variants retained | Plan/data identities and full results | NOT_RUN / not assessed by unit review |
| P7 | Untouched holdout and prospective paper/demo plan | Their separately frozen criteria resolved; no implicit live activation | Holdout/forward manifests and receipts | NOT_RUN |
For C, request next
Complete source/includes and permitted dependencies; exact specification/risk and input set; actual MetaEditor build/compiler log; broker symbol/timeframe/model/dates/costs/history completeness; original native runtime receipts, then any historical/holdout/forward evidence. Until provided, keep the EA-readiness decision INCONCLUSIVE. No order, alert, terminal action or live capital recommendation is authorized by this report.
Rights and claims
This bundled worked example is protected skill context, not a standalone buyer export. An independently generated review of a buyer's own supplied code is a separate output, subject to third-party rights. No guaranteed defect detection, comparative benchmark benefit, universal non-repainting, profit, prop-firm compliance or independent audit certification is claimed.
Actual product validation, 2026-10-21
- Package/context QA: 11 unittest tests PASS; exact seven-file manifest, deterministic ZIP, byte equality, relative context and negative path/missing/extra-file checks.
- Skill frontmatter validator: PASS. Scope-only1.0.1 keeps example source unchanged; earlier independent behavior evidence is historical, not a fresh independent trial of these narrowed instructions.
- Separate assistant trial read the skill/protocols/input units but not the worked answer, expected fixture file or earlier QA. A met 11/11 selected independently written PowerShell/.NET mirror expectations, with no supported finding in its bounded unit. B failed 3/11, including the supported HIGH upward-minimum clamp violation. C remained INCONCLUSIVE. Reports and source identities are retained separately.
- This is local arithmetic and assistant behavior evidence, not native MQL execution, exhaustive detection, external certification, a paired benchmark or demonstrated improvement over an unassisted baseline.
- Native MT5/Pine compilation, chart/runtime, historical, holdout and forward checks: NOT_RUN for this source-review/planning demonstration. Complete EAs and broker integration are outside the supplied units.
- Paid-draft live connector retrieval is unavailable before publication; live buyer-host and multi-host delivery: NOT_RUN. Local context loading is not a live delivery pass.
- No original strategy/risk changes, broker orders, alerts, account connections or live capital actions. Marketplace submission is separate from this demo and grants no trading authority.
Connects securely to your tools. The creator never sees your data.
About this skill
Evidence-led trading-code review, not a safety certificate
Review supplied MQL5 or TradingView Pine source against an explicit behavior and risk contract. Produce source-bound findings, justified no-findings dispositions, a claim/evidence ledger and a prioritized demo-only validation handoff. Missing evidence stays missing: a compiler claim or profitable backtest is not promoted into native safety or live readiness.
What you receive
- Finding location, severity, trigger, expected versus actual behavior, impact and minimal remedy proposal.
- Clear distinction between observed results, source reasoning, arithmetic mirrors and proposed tests.
- A reproducible validation matrix with prerequisites, frozen settings, pass criteria, artifacts and actual status.
- Honest source identities, partial-scope limits and independent versus self-review labels; no defect quota.
Agent-loaded supporting context
Seven files: SKILL.md, README.md, references/review-protocol.md, references/validation-plan.md, examples/review-inputs.md, examples/worked-review.md and examples/fixtures.json. Examples include two complete bounded MQL5-style arithmetic functions, a deliberately defective minimum clamp and a missing-evidence case. They are not standalone EAs. Relative Markdown/JSON context is loaded by your compatible Agensi-connected assistant, not offered as a buyer ZIP download.
$5 one-time scope
An assistant workflow and supporting context, not a compiler, hosted scanner, account manager, unlimited bespoke development, response SLA or lifetime-update promise. Reproducible package/use issues are in scope for support; refunds follow Agensi terms. Independently generated reports for your own supplied code are separate outputs subject to third-party rights. Bundled examples, package context and substantial reproductions remain protected and are not standalone exports. No universal licence clearance is claimed.
Safety and limitations
Demo/testing only. No grid, martingale, averaging, pyramiding or risk increases. Reviewing source does not authorize running submitted code, changing strategy/risk, connecting accounts or placing orders. Only MT5 and TradingView source-review/planning are supported, not native certification. Every actual product needs its own native and later economic/forward gates. No guaranteed bug detection, profitability, prop-firm compliance or live-readiness claim.
Actual validation
11 local package/context tests and skill-format validation PASS. A separate assistant trial, without reading the worked answer/expected fixtures/prior QA, identified the seeded HIGH clamp defect, gave a bounded no-findings disposition for A and left missing-source C inconclusive. A matched 11/11 selected independent arithmetic-mirror cases; B failed 3/11 as expected of the defective fixture. These are not native MQL executions, exhaustive tests, external certification or a paired baseline benchmark. Native MT5/Pine compilation/runtime and market-performance gates were not run for this review/planning demo. Paid unpublished drafts cannot be live-connector-tested before publication; live buyer-host and multi-host delivery remain unverified, not passed.
Source-review/planning workflow, not compiler/scanner/certification. Selected local package and arithmetic checks only; no guaranteed defect detection or paired benchmark. Native compile/runtime, historical, holdout and forward results NOT_RUN for this demo. No whole-EA acceptance from a pure arithmetic unit. Paid-draft live connector retrieval unavailable; buyer/multi-host delivery NOT_RUN. Bundled examples and substantial copies protected; independent reports subject to third-party rights. Demo-only; no risk or strategy alteration.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 27 days ago
Permissions
Read only authorized source, evidence and skill context. Optional writes save independently generated buyer reports, not protected skill/example exports. Optional browser/network reads public official documentation only. No Terminal/Shell or environment-variable access required; no execution of supplied code, credential access, orders, account changes, spending or publication authority.