- Home
- Skills
- Trading & Investing
- Trading Bot No-Trade Debugger
Trading Bot No-Trade Debugger
Find where your trading bot stops between signal and order. Audit rejection counters, SIM/LIVE gates, AI outages and execution evidence using local logs, with an offline JSONL helper and a reproducible demo.
$9
Works with the AI Agents you already use
Trading Bot No-Trade Debugger
Example session with this skill installed
My bot scans symbols but opens no live trades. Audit the supplied examples/cycles.jsonl, compare CRYPTO and VN_STOCK, identify the next checks, and tell me whether these logs prove a bug. Keep the diagnosis offline.
- Read your context and instructions
- Compiled the trading bot no-trade
Synthetic fixture audit
This is a reproducible software demo, not a production trading result. I ran the included helper on the three supplied cycle summaries. Its complete output follows.
Trading bot evidence summary
Input: 3 nonempty lines; 3 accepted cycle summaries.
Skipped: duplicate=0, conflict=0, malformed=0, unsupported=0, invalid=0, encrypted=0.
Coverage warnings: missingTimestamp=0, otherAsset=0.
CRYPTO
Cycles: 2. Observed timestamps: 2026-01-15T02:00:00.000Z to 2026-01-15T02:05:00.000Z.
| Counter | Total occurrences | Cycles reporting |
|---|---|---|
| scanned | 200 | 2/2 |
| weak | 75 | 2/2 |
| vol | 45 | 2/2 |
| setup | 25 | 2/2 |
| simOk | 55 | 2/2 |
| liveGate | 38 | 2/2 |
| aiVeto | 3 | 2/2 |
| risk | 7 | 2/2 |
| matchedSim | 11 | 2/2 |
| matchedLive | 0 | 2/2 |
| aiHardVeto | 3 | 2/2 |
Inspect nonzero blockers: weak=75, vol=45, liveGate=38, setup=25, risk=7, aiVeto=3.
VN_STOCK
Cycles: 1. Observed timestamps: 2026-01-15T02:05:00.000Z to 2026-01-15T02:05:00.000Z.
| Counter | Total occurrences | Cycles reporting |
|---|---|---|
| scanned | 50 | 1/1 |
| weak | 30 | 1/1 |
| setup | 12 | 1/1 |
| matchedSim | 2 | 1/1 |
| matchedLive | 0 | 1/1 |
Inspect nonzero blockers: weak=30, setup=12.
Interpretation limits
- Counter totals can overlap; do not subtract rejects or infer conversion rates.
- Missing counters are unknown; compare only metrics with matching coverage.
- Blocker ranking is an inspection priority, not proof of a defect or root cause.
- matchedLive is a pipeline match, not proof of submission, acknowledgement or fill.
- No raw reason text, symbols, account IDs or credentials are included.
Diagnosis
The CRYPTO input reports 11 SIM matches and zero LIVE matches over two observed cycles. Live-gate rejection events appear 38 times, but weak/volume/setup counters also appear; the counters are not a disjoint funnel. For VN_STOCK, volume, live-gate and risk counters were not supplied, so they remain unknown.
| Finding | Certainty | Evidence | Next check |
|---|---|---|---|
| CRYPTO SIM matching occurred while LIVE matching was zero | Confirmed for this fixture | examples/cycles.jsonl:1-2 | Compare SIM/LIVE predicates and package modes at the actual call sites |
| LIVE eligibility is worth inspecting | Likely investigation path; no defect proven | liveGate=38 across the two CRYPTO rows | Trace one rejected candidate with its gate reason and input values |
| VN_STOCK blocker detail is incomplete | Confirmed coverage limit | examples/cycles.jsonl:3 | Obtain the same-window gate summaries or relevant code branch |
| Broker submission and fill status | Unknown | No broker events were supplied | Correlate intent, submission, acknowledgement and fill evidence |
No code change is justified from these aggregate logs alone. In particular, lowering thresholds would be a strategy-policy change without evidence of a software defect. The next useful input is one candidate's rejection trace and the branch that produced it. The helper execution and totals above were verified; no broker or trading application was started.
Connects securely to your tools. The creator never sees your data.
About this skill
Signals appear. Orders do not. Find the missing step.
For developers running automated trading pipelines, a quiet bot can mean valid filtering, broken data, the wrong execution mode or a failed handoff to the broker. This skill helps your agent distinguish them using the logs and source code you supply.
What it delivers
- A map of the observed path from scan and setup filters through risk, AI confirmation, package matching and broker evidence.
- An offline JSONL summary grouped by asset, with missing metrics, duplicates, malformed rows and conflicting snapshots made explicit.
- Ranked findings tied to file/line evidence, with confirmed facts separated from hypotheses.
- A small repair plan and an offline regression check when the evidence supports a software fix.
Built from a real pipeline
Developed by inspecting OMNI DUCK's AutoDuck funnel and audit workflow, then packaged as independent instructions and a new dependency-free helper. It does not require OMNI DUCK, MongoDB, a data subscription or an exchange account. AutoDuck-shaped cycle logs are supported directly; another bot needs the documented cycle schema or manual mapping.
Included
One SKILL.md, a Node.js helper, three focused references, an incident report template and a synthetic JSONL demo. The helper reads one user-selected local file and prints a report; it makes no network calls. It omits arbitrary log text, account IDs and credentials from output.
Clear boundaries
Counter totals can overlap. A pipeline match does not prove an order submission or a fill. This skill does not weaken trading filters, place orders, restart your bot or promise investment returns. Plain text logs outside the schema require manual review. Encrypted logs require an already authorized plaintext export.
The helper passed 12 automated tests on Node 24.15.0 / Windows. Node 18+ is the compatibility target; this release has not been independently tested across every agent. Agent reasoning quality and the completeness of your evidence still affect the diagnosis.
Cycle-summary JSONL only for automated aggregation, up to 25 MiB per file and 1 MiB per line. Other formats require manual mapping. Encrypted records are skipped. Missing counters remain unknown. Multiple processes at the same asset/timestamp may conflict. No execution/fill or strategy-profitability guarantees. Cross-agent behavior has not been independently verified.
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 22 days ago
Requires
Permissions
Reads explicitly selected local logs and relevant source excerpts. The helper writes only to stdout and makes no network calls. An agent may write a temporary helper when delivered via MCP, an optional report, or a user-requested patch. No environment variables, account connection or broker credentials are required.