Desk Post-Trade Review: A Skill for Smarter Trading Agents

Where this fits

You've built an agent that sends orders, tracks positions, and closes trades. But when the dust settles, you have no structured way to answer "how did that go?" — the agent's memory is scattered across chat logs and raw exchange data. The desk-post-trade-review skill gives your agent a repeatable procedure for journaling every action and reviewing each trade, separating process quality from outcome luck.

Why agents benefit

• Process vs. outcome separation: The review grades how the trade was executed (risk checks, approval flow, order timing) independently of P&L, so the agent learns from good process even when the market moves against it.
• Execution cost measurement: It captures slippage, fees, and fill quality from the exchange record, turning vague "it filled okay" into quantifiable metrics.
• One repeatable finding: Each review produces a single actionable insight, preventing the agent from drowning in noise and focusing on one improvement per trade.
• Weekly desk review: The skill includes a routine that aggregates daily journals into a weekly summary, helping the agent spot patterns over time.
• Exchange-first, chat-second: It pulls data from the exchange record before considering user messages, ensuring the review is grounded in facts, not just conversation.

Practical scenarios

After a trade closes

Your agent just closed an ETH-PERP long. Instead of a free-form postmortem, it runs the review: logs the entry/exit, calculates execution cost, grades process steps (risk check, approval, order type), and writes one finding like "stop was placed 0.5% too tight — widen next time."

Weekly desk routine

Every Friday, the agent compiles the week's journal, identifies recurring process failures, and produces a desk review. This becomes the input for adjusting risk limits or order routing logic.

User asks "how did that go?"

When a user asks about a specific trade, the agent retrieves the journal entry and review, presenting the process grade and execution cost alongside the outcome. It doesn't guess — it reads from the structured journal.

Add it to your agent workflow

Install the skill via:

npx skills add galleonlabs/hypergrok-trading-desk --skill desk-post-trade-review

Then invoke it after any send or when a trade closes. For example, in your agent's post-trade handler:

await agent.skill('desk-post-trade-review', { tradeId: 'HG-20260816-01', action: 'review' });

The skill will append to the daily journal and return the review findings.

Compare before adopting

Before committing, compare against alternatives like manual journaling scripts or generic note-taking skills. Check the skill's maintenance freshness (last commit, open issues), its quality signals (stars, usage examples), and whether it fits your workflow — it's designed for trading desks with a specific journal structure. Also consider if you need the weekly review aggregation or just per-trade analysis.

Why it is worth tracking

With 35 stars and an MIT license, this skill shows early traction but is still niche. Its strength lies in the disciplined process it enforces — something most agent frameworks lack. Track it if you're building trading agents that need to improve from experience, and evaluate it again as the author updates it for new exchange APIs or agent frameworks. For a broader view of similar skills, see the OpenAgentSkill listing.