@delta-infra/cli 1.0.76 → 1.0.78

package/skills/lammps/references/endpoints.md

@@ -1,61 +1,61 @@
-# LAMMPS Endpoint Reference
-
-All paths are relative to `http://111.2.199.31:52317/api/v1`.
-
-## Health
-
-`GET /chem/lammps/health`
-
-Key response fields in `data`: `status`, `service`, `lammps_available`,
-`metadata`, and `limits`.
-
-Use this before long simulations. A healthy service should report
-`data.status == "ok"` and `data.lammps_available == true`.
-
-## Built-In Lennard-Jones Example
-
-`GET /chem/lammps/examples/lj-melt`
-
-Returns a complete `request` payload for `/chem/lammps/run`, plus workflow notes
-showing where pymatgen and GSAS-II fit around LAMMPS.
-
-## Run
-
-`POST /chem/lammps/run`
-
-```json
-{
-  "input_script": "<LAMMPS input script>",
-  "files": [
-    {"path": "system.data", "content": "<LAMMPS data file text>", "encoding": "text"}
-  ],
-  "timeout_seconds": 30,
-  "max_thermo_rows": 1000,
-  "log_tail_lines": 200,
-  "output_files": ["dump.lammpstrj", "final.data"],
-  "include_log": true,
-  "include_stdout": false,
-  "include_stderr": false
-}
-```
-
-`files` are written into a fresh job working directory before LAMMPS starts.
-`encoding` can be `text` or `base64`.
-
-The response includes parsed thermo rows, the final thermo row, atom count when
-detectable, log tail, warnings, and requested output files. Text output files
-return `content`; binary output files return `content_base64`.
-
-Notes:
-
-- Use relative file paths only.
-- The service rejects unsafe commands such as `shell`, `python`, `plugin`, and
-  `jump` by default.
-- `include <relative-file>` is allowed when the included file is supplied in
-  `files`; absolute paths, parent-directory traversal, URLs, and `jump` loops
-  are rejected.
-- Dump/data products are returned as raw requested output files, not structured
-  parsed atom-frame objects.
-- The gateway is synchronous; keep interactive runs short.
-- Force-field choice is caller responsibility. The service executes explicit
-  LAMMPS input and does not infer potentials from formulas or structures.
+# LAMMPS Endpoint Reference
+
+All paths are relative to `http://111.2.199.31:52317/api/v1`.
+
+## Health
+
+`GET /chem/lammps/health`
+
+Key response fields in `data`: `status`, `service`, `lammps_available`,
+`metadata`, and `limits`.
+
+Use this before long simulations. A healthy service should report
+`data.status == "ok"` and `data.lammps_available == true`.
+
+## Built-In Lennard-Jones Example
+
+`GET /chem/lammps/examples/lj-melt`
+
+Returns a complete `request` payload for `/chem/lammps/run`, plus workflow notes
+showing where pymatgen and GSAS-II fit around LAMMPS.
+
+## Run
+
+`POST /chem/lammps/run`
+
+```json
+{
+  "input_script": "<LAMMPS input script>",
+  "files": [
+    {"path": "system.data", "content": "<LAMMPS data file text>", "encoding": "text"}
+  ],
+  "timeout_seconds": 30,
+  "max_thermo_rows": 1000,
+  "log_tail_lines": 200,
+  "output_files": ["dump.lammpstrj", "final.data"],
+  "include_log": true,
+  "include_stdout": false,
+  "include_stderr": false
+}
+```
+
+`files` are written into a fresh job working directory before LAMMPS starts.
+`encoding` can be `text` or `base64`.
+
+The response includes parsed thermo rows, the final thermo row, atom count when
+detectable, log tail, warnings, and requested output files. Text output files
+return `content`; binary output files return `content_base64`.
+
+Notes:
+
+- Use relative file paths only.
+- The service rejects unsafe commands such as `shell`, `python`, `plugin`, and
+  `jump` by default.
+- `include <relative-file>` is allowed when the included file is supplied in
+  `files`; absolute paths, parent-directory traversal, URLs, and `jump` loops
+  are rejected.
+- Dump/data products are returned as raw requested output files, not structured
+  parsed atom-frame objects.
+- The gateway is synchronous; keep interactive runs short.
+- Force-field choice is caller responsibility. The service executes explicit
+  LAMMPS input and does not infer potentials from formulas or structures.

package/skills/ldm-bo/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: ldm-bo
 description: 当用户要求对小分子 SMILES 使用 LDM-BO/PDF2Dock、Vina+NN、ReaSyn，生成下一条小分子候选或运行分子 BO trajectory 时使用。不得用于抗体、CDRH3、AntBO 或 /biology/antbo 场景；这些场景必须路由 antbo-service。
 metadata:
@@ -7,29 +7,29 @@ metadata:
     - python_repl
     - step_finish
 ---
-
-# 小分子 LDM-BO 服务
-
-## 强制规则
-
-- 只用 `python_repl` 执行 `$SKILLS_ROOT/delta-science/scripts/invoke.py`，
-  tool=`ldm-bo`；禁止 Bash、路径探测、字面 CLI、直接 HTTP 和本地 PDF2Dock。
-- `python_repl` 工具参数只能包含 `code`，禁止 `deps/dependencies/packages`。
-- 业务结果直接位于 `r["native"]`，禁止 `native.data`。
-- provider、模型路径、Vina/ReaSyn/GPU/LLM 配置由服务端注入，不放入请求 JSON。
-- 抗体/CDRH3 即使出现 “LDM” 也禁止调用本 Skill。
-
-## operation 选择
-
-- `health`：检查 PDF2Dock/Vina/ReaSyn/NN/LLM 等部署依赖，无 body。
-- `recommend`：`method`、小分子 `pool`、真实
-  `history:[{smiles,scores}]`、`batch_size`。
-- `trajectory`：保留连字符键 `seed-smiles`、`num-evaluations`、`batch-size`，以及
-  method、seed、objective；有界 smoke 使用 `bo-tanimoto` 和很小 evaluation 数。
-
-详细最小 payload 和 native 响应见 `references/endpoints.md`。缺 Vina、ReaSyn、模型或
-LLM 配置时原样报告后端错误，不切换方法或本地生成候选。
-
-通过 wrapper 调用一次并显式打印有界 native 投影；成功后立即
-`step_finish(status="done", summary=<同一 JSON>)`，失败则
-`step_finish(status="failed", summary=<原始错误短摘要>)`。禁止重试和后续工具。
+
+# 小分子 LDM-BO 服务
+
+## 强制规则
+
+- 只用 `python_repl` 执行 `$SKILLS_ROOT/delta-science/scripts/invoke.py`，
+  tool=`ldm-bo`；禁止 Bash、路径探测、字面 CLI、直接 HTTP 和本地 PDF2Dock。
+- `python_repl` 工具参数只能包含 `code`，禁止 `deps/dependencies/packages`。
+- 业务结果直接位于 `r["native"]`，禁止 `native.data`。
+- provider、模型路径、Vina/ReaSyn/GPU/LLM 配置由服务端注入，不放入请求 JSON。
+- 抗体/CDRH3 即使出现 “LDM” 也禁止调用本 Skill。
+
+## operation 选择
+
+- `health`：检查 PDF2Dock/Vina/ReaSyn/NN/LLM 等部署依赖，无 body。
+- `recommend`：`method`、小分子 `pool`、真实
+  `history:[{smiles,scores}]`、`batch_size`。
+- `trajectory`：保留连字符键 `seed-smiles`、`num-evaluations`、`batch-size`，以及
+  method、seed、objective；有界 smoke 使用 `bo-tanimoto` 和很小 evaluation 数。
+
+详细最小 payload 和 native 响应见 `references/endpoints.md`。缺 Vina、ReaSyn、模型或
+LLM 配置时原样报告后端错误，不切换方法或本地生成候选。
+
+通过 wrapper 调用一次并显式打印有界 native 投影；成功后立即
+`step_finish(status="done", summary=<同一 JSON>)`，失败则
+`step_finish(status="failed", summary=<原始错误短摘要>)`。禁止重试和后续工具。

package/skills/ldm-bo/agents/openai.yaml

@@ -1,4 +1,4 @@
-interface:
-  display_name: "LDM-BO 小分子优化"
-  short_description: "通过 Delta CLI 生成小分子候选或执行分子优化轨迹"
-  default_prompt: "使用 $ldm-bo 为这个小分子任务推荐下一条候选。"
+interface:
+  display_name: "LDM-BO 小分子优化"
+  short_description: "通过 Delta CLI 生成小分子候选或执行分子优化轨迹"
+  default_prompt: "使用 $ldm-bo 为这个小分子任务推荐下一条候选。"

package/skills/ldm-bo/references/endpoints.md

@@ -1,127 +1,127 @@
-# LDM-BO API Endpoint Reference
-
-All paths are relative to `http://111.2.199.31:52317/api/v1`.
-
-This reference is for using the exposed HTTP API. It is not a local PDF2Dock
-runner.
-
-## Health
-
-`GET /chem/ldm-bo/health`
-
-The response reports whether the adapter can see the mounted PDF2Dock
-interpreter, `bo_api.py`, Vina binary, ReaSyn repo/interpreter/checkpoints, NN
-model files, cache directory, GPU settings, and LLM configuration. Missing LLM
-settings degrade health because LDM methods need them.
-
-## One-Step Recommend
-
-`POST /chem/ldm-bo/recommend`
-
-Wraps `bo_api.recommend_next_smiles`.
-
-Typical payload:
-
-```json
-{
-  "method": "bo-tanimoto",
-  "pool": ["CCO", "CCN", "CCC"],
-  "history": [
-    {"smiles": "CCO", "scores": [-7.1]},
-    {"smiles": "CCN", "scores": [-6.8]}
-  ],
-  "batch_size": 1
-}
-```
-
-Typical native response inside gateway `data`:
-
-```json
-{
-  "recommendations": ["CCC"],
-  "acquisition_values": [0.12],
-  "n_objectives": 1
-}
-```
-
-LDM methods may include an `llm` diagnostics object.
-
-## Full Trajectory
-
-`POST /chem/ldm-bo/trajectory`
-
-Wraps `bo_api.run_search_trajectory`.
-
-Small smoke payload:
-
-```json
-{
-  "method": "bo-tanimoto",
-  "seed": 42,
-  "seed-smiles": "CCO,CCN,CCC",
-  "num-evaluations": 4,
-  "batch-size": 1,
-  "objective": "vina+nn"
-}
-```
-
-Typical native response inside gateway `data`:
-
-```json
-{
-  "config": {},
-  "history": [],
-  "summary": {}
-}
-```
-
-LDM methods may include `llm_trajectory`.
-
-## Error Handling
-
-The service maps native `bo_api` error JSON to an upstream failure. The gateway
-returns a standard error envelope:
-
-```json
-{
-  "code": 1,
-  "message": "AutoDock Vina executable not found...",
-  "data": {
-    "endpoint": "/v1/trajectory",
-    "service_url": "http://ldm-bo-service:8895",
-    "ldm_bo_response": {
-      "error": "...",
-      "error_type": "ValueError",
-      "traceback": "Traceback ..."
-    }
-  }
-}
-```
-
-Common causes are missing Vina, missing ReaSyn, wrong Python interpreter, CUDA
-device mismatch, missing LLM credentials for LDM methods, or an unwritable Vina
-cache directory.
-
-## Provider Settings
-
-Do not send provider settings in request JSON. The adapter reads these from
-deployment config and passes them as Python kwargs to `bo_api`:
-
-```text
-vina_bin
-vina_cache_dir
-vina_max_workers
-gp_device
-reasyn_repo
-reasyn_python_bin
-reasyn_model_path
-reasyn_devices
-nn_model_path
-nn_metadata_path
-llm_model
-llm_base_url
-llm_api_key
-```
-
-The wrapper intentionally preserves user payloads and lets `bo_api` own native
-defaults and schema behavior.
+# LDM-BO API Endpoint Reference
+
+All paths are relative to `http://111.2.199.31:52317/api/v1`.
+
+This reference is for using the exposed HTTP API. It is not a local PDF2Dock
+runner.
+
+## Health
+
+`GET /chem/ldm-bo/health`
+
+The response reports whether the adapter can see the mounted PDF2Dock
+interpreter, `bo_api.py`, Vina binary, ReaSyn repo/interpreter/checkpoints, NN
+model files, cache directory, GPU settings, and LLM configuration. Missing LLM
+settings degrade health because LDM methods need them.
+
+## One-Step Recommend
+
+`POST /chem/ldm-bo/recommend`
+
+Wraps `bo_api.recommend_next_smiles`.
+
+Typical payload:
+
+```json
+{
+  "method": "bo-tanimoto",
+  "pool": ["CCO", "CCN", "CCC"],
+  "history": [
+    {"smiles": "CCO", "scores": [-7.1]},
+    {"smiles": "CCN", "scores": [-6.8]}
+  ],
+  "batch_size": 1
+}
+```
+
+Typical native response inside gateway `data`:
+
+```json
+{
+  "recommendations": ["CCC"],
+  "acquisition_values": [0.12],
+  "n_objectives": 1
+}
+```
+
+LDM methods may include an `llm` diagnostics object.
+
+## Full Trajectory
+
+`POST /chem/ldm-bo/trajectory`
+
+Wraps `bo_api.run_search_trajectory`.
+
+Small smoke payload:
+
+```json
+{
+  "method": "bo-tanimoto",
+  "seed": 42,
+  "seed-smiles": "CCO,CCN,CCC",
+  "num-evaluations": 4,
+  "batch-size": 1,
+  "objective": "vina+nn"
+}
+```
+
+Typical native response inside gateway `data`:
+
+```json
+{
+  "config": {},
+  "history": [],
+  "summary": {}
+}
+```
+
+LDM methods may include `llm_trajectory`.
+
+## Error Handling
+
+The service maps native `bo_api` error JSON to an upstream failure. The gateway
+returns a standard error envelope:
+
+```json
+{
+  "code": 1,
+  "message": "AutoDock Vina executable not found...",
+  "data": {
+    "endpoint": "/v1/trajectory",
+    "service_url": "http://ldm-bo-service:8895",
+    "ldm_bo_response": {
+      "error": "...",
+      "error_type": "ValueError",
+      "traceback": "Traceback ..."
+    }
+  }
+}
+```
+
+Common causes are missing Vina, missing ReaSyn, wrong Python interpreter, CUDA
+device mismatch, missing LLM credentials for LDM methods, or an unwritable Vina
+cache directory.
+
+## Provider Settings
+
+Do not send provider settings in request JSON. The adapter reads these from
+deployment config and passes them as Python kwargs to `bo_api`:
+
+```text
+vina_bin
+vina_cache_dir
+vina_max_workers
+gp_device
+reasyn_repo
+reasyn_python_bin
+reasyn_model_path
+reasyn_devices
+nn_model_path
+nn_metadata_path
+llm_model
+llm_base_url
+llm_api_key
+```
+
+The wrapper intentionally preserves user payloads and lets `bo_api` own native
+defaults and schema behavior.

package/skills/pubchem/SKILL.md

@@ -1,4 +1,4 @@
----
+---
 name: pubchem
 description: 当用户给出化学品、分子、药物名称并询问 CID、同义词、分子式、分子量、SMILES、InChI、IUPAC 名、XLogP、TPSA 或供受体数量时使用；也用于在 RDKit 前批量把名称解析为 SMILES。所有实时请求只走 Delta CLI wrapper，并以当前 WORKSPACE_ROOT 下的有界 JSON 作为权威交接产物。
 metadata:
@@ -7,87 +7,87 @@ metadata:
     - python_repl
     - step_finish
 ---
-
-# PubChem 服务
-
-## 强制执行规则
-
-1. 第一个且唯一实时执行工具是 `python_repl`，直接执行
-   `$SKILLS_ROOT/delta-science/scripts/invoke.py`。
-   调用 `python_repl` 时工具参数对象只能包含 `code`；禁止传入 `deps`、
-   `dependencies`、`packages` 或其他参数。
-2. `SKILLS_ROOT` 只通过 `os.environ["SKILLS_ROOT"]` 读取；禁止 Bash、
-   `list_dir/read_file/which/where/ls`、wrapper 检查和路径探测。
-3. 禁止执行字面命令 `CLI/cli/delta-cli`，禁止 curl、requests、httpx、浏览器、
-   PUG-REST 和本地替代。
-4. wrapper 已解包所有 envelope，业务结果直接位于 `r["native"]`；禁止
-   `native.data` 和递归找值。
-5. 不先调用 health/schema，不为“验证”重复同一请求。
-
-## 单化合物默认模板
-
-```python
-import json, os, subprocess, sys
-from pathlib import Path
-wrapper=os.path.join(os.environ["SKILLS_ROOT"],"delta-science","scripts","invoke.py")
-payload={"identifier":"aspirin","namespace":"name"}
-p=subprocess.run([sys.executable,wrapper,"--tool","pubchem","--endpoint","compound-summary",
-                  "--data-json",json.dumps(payload)],capture_output=True,text=True,
-                 encoding="utf-8",errors="replace",shell=False,timeout=150)
-if p.returncode != 0: raise RuntimeError(p.stderr or p.stdout)
-r=json.loads(p.stdout)
-if r.get("ok") is not True or r.get("transport")!="delta-cli": raise RuntimeError(r)
-n=r["native"]; x=n["properties"]
-out={"transport":r["transport"],"endpoint":r["endpoint"],"valid":n["valid"],
-     "cid":n.get("cid") or x.get("CID"),"canonical_smiles":x.get("CanonicalSMILES"),
-     "molecular_formula":x.get("MolecularFormula"),
-     "molecular_weight":x.get("MolecularWeight")}
-output_path=Path(os.environ["WORKSPACE_ROOT"])/"pubchem_result.json"
-output_path.write_text(json.dumps(out,ensure_ascii=False,indent=2),encoding="utf-8")
-print(json.dumps({"output_ref":str(output_path),"result":out},ensure_ascii=False))
-```
-
-`out` 只保留用户要求的字段。逐字复制返回值：不得改写等价 SMILES、给公式加排版、
-给数值加单位、添加 URL/链接/用途/机制或未请求属性。
-
-## operation 选择
-
-- 一个名称的基础信息：一次 `compound-summary`。
-- 两个及以上名称：一次 `compound-batch-summary`，body 只用
-  `{"identifiers":[...],"namespace":"name"}`。
-- 只解析 CID、只查属性或同义词：分别使用 `compound-resolve-cids`、
-  `compound-properties`、`compound-synonyms`。
-- 每个 operation 的准确情景、payload 和投影见
-  `references/operations.md`；复杂字段再看 `references/endpoints.md`。
-
-批量结果从 `native.results[*]` 读取，原标签是 `input`，化合物字段在嵌套
-`properties`。SMILES 回退顺序固定为：
-
-```text
-CanonicalSMILES || ConnectivitySMILES || SMILES || IsomericSMILES
-```
-
-无机材料化学式、组成和式量路由 pymatgen；PubChem not_found 原样保留，不编造 CID
-或本地补算。PubChem `XLogP` 与 RDKit `MolLogP` 不得互换。
-
-成功写入有界 JSON 后，用其中的原始字符串和数值构造无 Markdown、无单位的最终文本，
-并立即调用 `step_finish` 恰好一次。例如：
-
-```text
-【最终答案，必须逐字转发】
-CID: 2244
-规范 SMILES: CC(=O)OC1=CC=CC=C1C(=O)O
-分子式: C9H8O4
-分子量: 180.16
-```
-
-`status="done"`，`summary` 第一行必须逐字为 `【最终答案，必须逐字转发】`，其余行
-逐字使用当前 out；`output_ref` 必须是刚写入的绝对 `pubchem_result.json` 路径。
-不得使用示例固定值。缺少首行或 output_ref 时不得结束步骤。
-Skill 自身的 final_response 和外层 Memento 都必须逐字复制 summary，包括大小写和标点，
-不得改成 Markdown/LaTeX、另一种等价 SMILES、添加单位、链接或说明。失败时调用
-`step_finish(status="failed", summary=<原始错误短摘要>)`。调用后禁止自由文本和工具。
-
-具名 Skill 返回后，外层若需要核验只能对 `output_ref` 使用一次原子 `read_file`；禁止
-`python_repl`、`list_dir`、路径探测或再次调用 PubChem/Science。该文件只是当前会话的
-内部证据交接，不是用户要求的额外交付物。
+
+# PubChem 服务
+
+## 强制执行规则
+
+1. 第一个且唯一实时执行工具是 `python_repl`，直接执行
+   `$SKILLS_ROOT/delta-science/scripts/invoke.py`。
+   调用 `python_repl` 时工具参数对象只能包含 `code`；禁止传入 `deps`、
+   `dependencies`、`packages` 或其他参数。
+2. `SKILLS_ROOT` 只通过 `os.environ["SKILLS_ROOT"]` 读取；禁止 Bash、
+   `list_dir/read_file/which/where/ls`、wrapper 检查和路径探测。
+3. 禁止执行字面命令 `CLI/cli/delta-cli`，禁止 curl、requests、httpx、浏览器、
+   PUG-REST 和本地替代。
+4. wrapper 已解包所有 envelope，业务结果直接位于 `r["native"]`；禁止
+   `native.data` 和递归找值。
+5. 不先调用 health/schema，不为“验证”重复同一请求。
+
+## 单化合物默认模板
+
+```python
+import json, os, subprocess, sys
+from pathlib import Path
+wrapper=os.path.join(os.environ["SKILLS_ROOT"],"delta-science","scripts","invoke.py")
+payload={"identifier":"aspirin","namespace":"name"}
+p=subprocess.run([sys.executable,wrapper,"--tool","pubchem","--endpoint","compound-summary",
+                  "--data-json",json.dumps(payload)],capture_output=True,text=True,
+                 encoding="utf-8",errors="replace",shell=False,timeout=150)
+if p.returncode != 0: raise RuntimeError(p.stderr or p.stdout)
+r=json.loads(p.stdout)
+if r.get("ok") is not True or r.get("transport")!="delta-cli": raise RuntimeError(r)
+n=r["native"]; x=n["properties"]
+out={"transport":r["transport"],"endpoint":r["endpoint"],"valid":n["valid"],
+     "cid":n.get("cid") or x.get("CID"),"canonical_smiles":x.get("CanonicalSMILES"),
+     "molecular_formula":x.get("MolecularFormula"),
+     "molecular_weight":x.get("MolecularWeight")}
+output_path=Path(os.environ["WORKSPACE_ROOT"])/"pubchem_result.json"
+output_path.write_text(json.dumps(out,ensure_ascii=False,indent=2),encoding="utf-8")
+print(json.dumps({"output_ref":str(output_path),"result":out},ensure_ascii=False))
+```
+
+`out` 只保留用户要求的字段。逐字复制返回值：不得改写等价 SMILES、给公式加排版、
+给数值加单位、添加 URL/链接/用途/机制或未请求属性。
+
+## operation 选择
+
+- 一个名称的基础信息：一次 `compound-summary`。
+- 两个及以上名称：一次 `compound-batch-summary`，body 只用
+  `{"identifiers":[...],"namespace":"name"}`。
+- 只解析 CID、只查属性或同义词：分别使用 `compound-resolve-cids`、
+  `compound-properties`、`compound-synonyms`。
+- 每个 operation 的准确情景、payload 和投影见
+  `references/operations.md`；复杂字段再看 `references/endpoints.md`。
+
+批量结果从 `native.results[*]` 读取，原标签是 `input`，化合物字段在嵌套
+`properties`。SMILES 回退顺序固定为：
+
+```text
+CanonicalSMILES || ConnectivitySMILES || SMILES || IsomericSMILES
+```
+
+无机材料化学式、组成和式量路由 pymatgen；PubChem not_found 原样保留，不编造 CID
+或本地补算。PubChem `XLogP` 与 RDKit `MolLogP` 不得互换。
+
+成功写入有界 JSON 后，用其中的原始字符串和数值构造无 Markdown、无单位的最终文本，
+并立即调用 `step_finish` 恰好一次。例如：
+
+```text
+【最终答案，必须逐字转发】
+CID: 2244
+规范 SMILES: CC(=O)OC1=CC=CC=C1C(=O)O
+分子式: C9H8O4
+分子量: 180.16
+```
+
+`status="done"`，`summary` 第一行必须逐字为 `【最终答案，必须逐字转发】`，其余行
+逐字使用当前 out；`output_ref` 必须是刚写入的绝对 `pubchem_result.json` 路径。
+不得使用示例固定值。缺少首行或 output_ref 时不得结束步骤。
+Skill 自身的 final_response 和外层 Memento 都必须逐字复制 summary，包括大小写和标点，
+不得改成 Markdown/LaTeX、另一种等价 SMILES、添加单位、链接或说明。失败时调用
+`step_finish(status="failed", summary=<原始错误短摘要>)`。调用后禁止自由文本和工具。
+
+具名 Skill 返回后，外层若需要核验只能对 `output_ref` 使用一次原子 `read_file`；禁止
+`python_repl`、`list_dir`、路径探测或再次调用 PubChem/Science。该文件只是当前会话的
+内部证据交接，不是用户要求的额外交付物。

package/skills/pubchem/agents/openai.yaml

@@ -1,4 +1,4 @@
-interface:
-  display_name: "PubChem 化合物查询"
-  short_description: "通过 Delta CLI 查询化合物身份、结构与公共性质"
-  default_prompt: "使用 $pubchem 查询这个化合物的基础身份和分子性质。"
+interface:
+  display_name: "PubChem 化合物查询"
+  short_description: "通过 Delta CLI 查询化合物身份、结构与公共性质"
+  default_prompt: "使用 $pubchem 查询这个化合物的基础身份和分子性质。"