desic-okx-agent 0.3.1 → 0.3.3

package/skills/okx-strategy-research/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: okx-strategy-research
-description: Write, validate, and backtest Python trading strategies against local OKX one-minute history. Use for strategy authoring, backtest execution, reading backtest results, and diagnosing local data coverage or gaps.
+description: Write, validate, and backtest Python trading strategies against local OKX one-minute history. Use for strategy authoring, backtest execution, reading results, generating or opening detailed standalone HTML reports, visually comparing runs, and diagnosing local data coverage or gaps.
 ---
 
 # OKX Strategy Research
@@ -40,6 +40,11 @@ names directly; do not probe for spellings.
    quoting two reports in sequence. Read its `warnings` first and repeat them: two
    runs over different data, instruments, windows, or costs are not two answers to
    one question, and that is invisible in the metrics alone.
+9. When the user asks for a detailed report, visual report, HTML report, charts,
+   or to open/view a report (including Chinese requests such as "详细报告",
+   "HTML 报告", "打开报告", or "查看图表"), generate it immediately with
+   `strategy_open_report` or `strategy_open_comparison`. Do not replace that action
+   with paged data reads or merely tell the user which CLI command they could run.
 
 ## Where this leads
 
@@ -77,6 +82,55 @@ describe a backtest result as expected future return.
 When `marginExhausted` is true, say the run hit its simulated collateral limit
 and that this is a research risk boundary, not an OKX liquidation estimate.
 
+### Generate and open the detailed HTML report
+
+A completed backtest's detailed view is the standalone HTML report, not a dump of
+`strategy_get_run_equity` or `strategy_get_run_trades`. It contains the interactive
+equity and drawdown curves, execution orbit, return spectrum, monthly performance,
+trade distribution, assumptions, and full trade ledger. All assets are inline, so
+the file remains private and works offline.
+
+When a completed run is available, or whenever the user asks to open or view its
+detailed/visual/HTML report, call the tool instead of only describing it:
+
+```text
+strategy_open_report { "runId": "bt_abc123" }
+```
+
+Omit `open` (or pass `"open": true`) to launch the operator's default browser.
+Use `"open": false` only when the user explicitly asks to generate the file
+without opening it. The result carries `data.file` and `data.opened`:
+
+- When `opened` is `true`, say the report was opened and still provide `file` so it
+  can be reopened or archived.
+- When `opened` is `false`, say the HTML was generated but the browser did not
+  launch, then provide the exact `file` path for the user to open locally. Do not
+  claim it opened.
+
+For two to six runs, generate the comparison observatory rather than opening
+several single-run files:
+
+```text
+strategy_open_comparison { "runIds": ["bt_abc123", "bt_def456"] }
+```
+
+It opens one normalized multi-curve report with the metrics matrix and assumption
+differences. Read `data.warnings` and repeat every warning in your own reply rather
+than leaving it only in the document. Different data, instruments, windows, or
+costs are not two answers to one question.
+
+If the MCP open tools are unavailable but the Desic CLI is available, the direct
+operator equivalents are:
+
+```bash
+desic-okx strategy report --run bt_abc123 --open
+desic-okx strategy compare --runs bt_abc123,bt_def456 --open
+```
+
+Do not open a report for a queued, running, failed, or cancelled run. Poll a queued
+or running backtest to a terminal status first; report failures and cancellations
+instead of producing a document without the result it exists to show.
+
 ## Judging a result
 
 A high trade count with a profit factor near or below 1 usually means fees

package/skills/okx-strategy-research/agents/openai.yaml

@@ -1,6 +1,6 @@
 interface:
   display_name: "OKX Strategy Research"
-  short_description: "Write, validate, and backtest Python strategies on local OKX history."
+  short_description: "Backtest OKX strategies and open detailed visual HTML reports."
   default_prompt: "Use $okx-strategy-research to backtest an EMA trend strategy on BTC-USDT-SWAP."
 dependencies:
   tools:

package/skills/okx-strategy-research/references/python-api.md

@@ -41,6 +41,13 @@ Always pass `lookback` when only a tail is needed. It costs the size of the wind
 asked for, whereas omitting it copies every bar seen so far on every call — on a
 long run that is the difference between a few seconds and a few minutes.
 
+A higher timeframe does not have to be declared on the request. The engine reads the
+source and aggregates whatever interval a `bars` call names, so writing
+`ctx.market.bars(id, "15m")` is enough. The run's `assumptions.intervals` reports what
+was actually aggregated, which is what to quote. An interval built at runtime —
+`bars(id, ctx.params["tf"])` — cannot be read from the source, so that case still needs
+`intervals` on the request.
+
 `Bar` fields: `openTimeMs`, `closeTimeMs`, `open`, `high`, `low`, `close`,
 `volume`, `confirmed`.
 

package/skills/okx-strategy-research/references/tools-and-data.md

@@ -30,9 +30,29 @@ There is no `bar` parameter on any data tool.
 | `strategy_get_run_trades` | Paged closed trades. |
 | `strategy_get_run_actions` | Paged decisions the strategy emitted. |
 | `strategy_compare_runs` | Two or more runs side by side, with the assumptions that differ and comparability warnings. |
+| `strategy_open_report` | Write a standalone HTML report for one run and open it in the operator's browser. |
+| `strategy_open_comparison` | The same for two or more runs, overlaying their equity curves. |
 | `strategy_cancel_run` | Cancel a queued or running run. |
 | `strategy_delete_run` | Delete a run and its stored series. Confirm with the user first. |
 
+Both open tools return their result under `data`: `data.file` is the absolute HTML
+path and `data.opened` says whether the default browser launched. Omit `open` to
+open by default; pass `open: false` only when the user asks to generate without
+opening. Always give the path. When `data.opened` is false, say the document was
+written but do not claim it opened. `strategy_open_comparison` also returns
+`data.warnings`, which belong in your reply and not only in the document.
+
+Use these calls for requests such as “detailed report”, “HTML report”, “open the
+report”, “详细报告”, “打开报告”, or “查看图表”:
+
+```text
+strategy_open_report { "runId": "bt_abc123" }
+strategy_open_comparison { "runIds": ["bt_abc123", "bt_def456"] }
+```
+
+The report tools collect all required pages internally. Do not page the full equity
+curve or trade ledger into the conversation before calling them.
+
 ## Check coverage before backtesting
 
 ```