dsh-plugin-subagent-bridge 0.15.2 → 0.17.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md CHANGED
@@ -145,13 +145,22 @@ Two measured facts:
145
145
  A worker that calls `ask_user_question` waits for a human. When the caller is another agent, nobody is there. So the bridge appends a standing instruction to every delegated task:
146
146
 
147
147
  ```
148
- (如果缺少必要信息或需要在多个方案里选择,直接采用最合理的默认做法继续做完,
149
- 不要把选择权抛回给用户、不要调用 ask_user_question 等回答;
150
- 在结果里用一行说明你假设了什么。)
148
+ (If information is missing, or a choice between options is required, pick the most
149
+ reasonable default, finish the task, and state the assumption in one line. Do not
150
+ hand the decision back to the user and do not call ask_user_question.)
151
+ ```
152
+
153
+ Set your own wording with `assumptionPolicy`:
154
+
155
+ ```yaml
156
+ config:
157
+ assumptionPolicy: "If information is missing, pick the most reasonable default, finish the task, and state the assumption in one line. Do not ask the user and do not call ask_user_question."
151
158
  ```
152
159
 
153
160
  On by default (`assumeDefaults: false` disables it). It is skipped when the prompt already mentions `ask_user_question`.
154
161
 
162
+ > The built-in default text is Chinese (`如果缺少必要信息……`), which every model in this host reads; the block above is that instruction in English. Set `assumptionPolicy` to send English verbatim.
163
+
155
164
  ### Questions (`pendingQuestion`)
156
165
 
157
166
  If a worker does ask, `dsh_wait` reports it:
@@ -215,6 +224,7 @@ Returns `{ ok, text, sessionId, durationMs }`. Runs `claude -p --output-format j
215
224
  | `maxResultChars` | `20000` | Cap on a single result payload (the head is kept) |
216
225
  | `taskTimeoutSeconds` | `300` | `dsh_task` wait cap (client may override, max 240) |
217
226
  | `assumeDefaults` | `true` | Append the decide-and-continue instruction |
227
+ | `assumptionPolicy` | *(empty)* | Replacement text for the instruction; empty uses the built-in Chinese default |
218
228
  | `workersFile` | `~/.dsh/bridge/workers.json` | Worker registry (`0600`) |
219
229
  | `maxHops` | `1` | Delegation depth limit |
220
230
  | `dispatchPerMinute` | `10` | Dispatches allowed per window |
@@ -297,6 +307,23 @@ Both are plain `SKILL.md` files: YAML frontmatter (`name` + `description`) plus
297
307
  node --test "test/*.test.mjs"
298
308
  ```
299
309
 
310
+ ```
311
+ index.js wiring: apply() + re-exports
312
+ src/constants.js versions, limits, binary candidates, policy text
313
+ src/utils.js JSON-RPC helpers, argument validation, sleep
314
+ src/settings.js defaults and config resolution
315
+ src/tools.js tool schemas
316
+ src/policy.js the standing instruction
317
+ src/questions.js ask_user_question takeover and reporting
318
+ src/sessions.js session reads, waits, model and workspace placement
319
+ src/guards.js depth and rate limits
320
+ src/registry.js worker registry persistence and capacity
321
+ src/http.js loopback server, JSON-RPC, tool dispatch
322
+ src/api.js the bridge API the tools call
323
+ src/runners/codex.js codex exec adapter
324
+ src/runners/claude.js claude -p adapter
325
+ ```
326
+
300
327
  83 tests: handshake and version negotiation, tool list and argument validation, dispatch call shape, result baselines, long polling, real HTTP round trips (404/405/202 and bearer), `codex exec` event parsing, binary resolution, success/failure/timeout runs, resume argument split, final-answer-only behaviour, `dsh_wait` terminal states, slot limits and release, followup rebaselining, registry persistence and restart recovery, `maxHops` and rate limiting, per-`cwd` budgets, workspace creation and reuse, `queue`/`steer` delivery and waiting, anchors for steered answers, Claude output parsing and resume arguments, `assumeDefaults` injection, `pendingQuestion` and `dsh_answer`.
301
328
 
302
329
  ## Related
package/README.zh-CN.md CHANGED
@@ -152,6 +152,13 @@ dsh_followup({ worker: "w1", message: "接下来只改 src 下的文件", mode:
152
152
 
153
153
  默认开启(`assumeDefaults: false` 可关闭)。prompt 里本来就提到 `ask_user_question` 时不追加。
154
154
 
155
+ 想换成自己的措辞——比如英文——用 `assumptionPolicy`:
156
+
157
+ ```yaml
158
+ config:
159
+ assumptionPolicy: "If information is missing, pick the most reasonable default, finish the task, and state the assumption in one line. Do not ask the user or call ask_user_question."
160
+ ```
161
+
155
162
  ### 问题上报
156
163
 
157
164
  worker 真的提问时,`dsh_wait` 会把问题带出来:
@@ -215,6 +222,7 @@ worker 真的提问时,`dsh_wait` 会把问题带出来:
215
222
  | `maxResultChars` | `20000` | 单次结果载荷上限(保留开头) |
216
223
  | `taskTimeoutSeconds` | `300` | `dsh_task` 等待上限(客户端可覆盖,最大 240) |
217
224
  | `assumeDefaults` | `true` | 追加「自行决策」指令 |
225
+ | `assumptionPolicy` | 空 | 替换指令正文;空表示用内置中文默认 |
218
226
  | `workersFile` | `~/.dsh/bridge/workers.json` | worker 注册表(`0600`) |
219
227
  | `maxHops` | `1` | 委派深度上限 |
220
228
  | `dispatchPerMinute` | `10` | 每个窗口允许的派发次数 |
@@ -297,6 +305,23 @@ mkdir -p ~/.dsh/skills && cp -R skills/delegate-to-external-agent ~/.dsh/skills
297
305
  node --test "test/*.test.mjs"
298
306
  ```
299
307
 
308
+ ```
309
+ index.js wiring: apply() + re-exports
310
+ src/constants.js versions, limits, binary candidates, policy text
311
+ src/utils.js JSON-RPC helpers, argument validation, sleep
312
+ src/settings.js defaults and config resolution
313
+ src/tools.js tool schemas
314
+ src/policy.js the standing instruction
315
+ src/questions.js ask_user_question takeover and reporting
316
+ src/sessions.js session reads, waits, model and workspace placement
317
+ src/guards.js depth and rate limits
318
+ src/registry.js worker registry persistence and capacity
319
+ src/http.js loopback server, JSON-RPC, tool dispatch
320
+ src/api.js the bridge API the tools call
321
+ src/runners/codex.js codex exec adapter
322
+ src/runners/claude.js claude -p adapter
323
+ ```
324
+
300
325
  83 项测试:握手与版本协商、工具清单与参数校验、派发调用形状、结果基线、长轮询、真实 HTTP 往返(404/405/202 与 bearer)、`codex exec` 事件解析、二进制解析、成功/失败/超时、resume 参数分叉、只回最终回答、`dsh_wait` 终态、名额上限与释放、followup 重基线、注册表持久化与重启恢复、`maxHops` 与速率限制、按 `cwd` 的预算、项目创建与复用、`queue`/`steer` 投递与等待、插话回答的锚点、Claude 输出解析与 resume 参数、`assumeDefaults` 注入、`pendingQuestion` 与 `dsh_answer`。
301
326
 
302
327
  ## 相关插件