@crewx/wi 0.1.10-rc.115 → 0.1.10-rc.117

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.
Files changed (2) hide show
  1. package/SKILL.md +60 -41
  2. package/package.json +3 -3
package/SKILL.md CHANGED
@@ -11,6 +11,23 @@ metadata:
11
11
  > **Core principle**: Context belongs in the file, not in your head.
12
12
  > An agent should be able to execute a WI without asking questions.
13
13
 
14
+ ## Command support
15
+
16
+ The dispatch logic in `packages/built-in/wi/cli.ts` is the source of truth for this table.
17
+
18
+ | Route | Status | Notes |
19
+ |--------|--------|-------|
20
+ | `usage` | Implemented | Shows this guide. |
21
+ | `create-batch` | Implemented | Creates gated draft WIs from batch JSON. |
22
+ | `board tick` | Implemented | Advances one approved board WI when gates pass. |
23
+ | `board digest` | Implemented | Sends and prints the overnight board summary. |
24
+ | `draft` | Planned / not implemented | Manual procedure below; not a runnable CLI route. |
25
+ | `issue` | Planned / not implemented | Manual procedure below; not a runnable CLI route. |
26
+ | `list` | Planned / not implemented | Manual procedure below; not a runnable CLI route. |
27
+ | `review` | Planned / not implemented | Manual procedure below; not a runnable CLI route. |
28
+
29
+ The four planned routes currently print `'<command>' is not yet implemented.` and return exit code `0`. Do not treat that exit code as proof that the requested operation succeeded.
30
+
14
31
  ## When to use
15
32
 
16
33
  - When delegating work to an agent
@@ -21,7 +38,7 @@ metadata:
21
38
 
22
39
  ---
23
40
 
24
- ## Commands
41
+ ## Implemented commands
25
42
 
26
43
  ### create-batch — Materialize a gated WI plan
27
44
 
@@ -86,15 +103,15 @@ Tick behavior:
86
103
  crewx wi board digest brd_release --json
87
104
  ```
88
105
 
89
- ### draft — Create a draft
106
+ ## Manual procedures
107
+
108
+ ### Manual procedure: draft — Create a draft
109
+
110
+ There is no implemented CLI route for `draft`; follow this procedure manually.
90
111
 
91
112
  Takes the user's intent and drafts a Work Instruction.
92
113
  The lead reads the design doc and fills in the details.
93
114
 
94
- ```bash
95
- crewx skill wi draft "task description"
96
- ```
97
-
98
115
  **Lead execution steps:**
99
116
 
100
117
  1. Check today's date + existing WI count → assign ID
@@ -104,20 +121,20 @@ crewx skill wi draft "task description"
104
121
 
105
122
  2. Read relevant section from design doc
106
123
  ```bash
107
- npx doc section docs/design.md "section name"
124
+ crewx doc section docs/design.md "section name"
108
125
  ```
109
126
 
110
127
  3. Identify affected files + **impact analysis**
111
128
  ```bash
112
129
  # Track symbol references
113
- npx skill tsserver find "symbolName"
114
- npx skill tsserver refs path/to/file.ts lineNumber
130
+ crewx skill tsserver find "symbolName"
131
+ crewx skill tsserver references path/to/file.ts lineNumber columnNumber
115
132
 
116
133
  # If API changed, find consumers
117
- npx search "endpointName" ./src/ui --glob="**/*.ts,**/*.tsx"
134
+ crewx search "endpointName" ./src/ui --glob="**/*.ts,**/*.tsx"
118
135
 
119
136
  # If types changed, find usages
120
- npx search "typeName" ./src --glob="**/*.ts,**/*.tsx"
137
+ crewx search "typeName" ./src --glob="**/*.ts,**/*.tsx"
121
138
  ```
122
139
 
123
140
  → Fill in `api_changed`, `schema_changed`, `consumers`, `risk_level`
@@ -143,39 +160,46 @@ crewx skill wi draft "task description"
143
160
 
144
161
  ---
145
162
 
146
- ### issue — Issue to an agent
163
+ ### Manual procedure: issue — Issue to an agent
147
164
 
148
- Deliver an approved Work Instruction to the assigned agent.
165
+ There is no implemented CLI route for `issue`; use the procedure below to issue an approved Work Instruction.
149
166
 
150
- ```bash
151
- crewx skill wi issue WI-20260312-001
152
- ```
167
+ Deliver an approved Work Instruction to the assigned agent.
153
168
 
154
169
  **Issue steps:**
155
170
 
156
- 1. Read WI file → check status (must be `approved`)
157
-
158
- 2. Check WI's `impl_agent` (`consultant` is for advisory — not an issue target)
171
+ 1. Read the WI file, confirm its status is `approved`, and identify its `impl_agent` before dispatching. (`consultant` is for advisory work, not an issue target.)
159
172
 
160
- 3. Check agent memory (any in-progress work?)
173
+ 2. Check agent memory (any in-progress work?)
161
174
  ```bash
162
- npx memory index {impl_agent}
175
+ crewx memory index {impl_agent}
163
176
  ```
164
177
 
165
- 4. Issue via `$CREWX_CLI x`:
178
+ 3. Invoke `crewx x` with the entire WI content in the prompt and `--thread="WI-ID"` (replace `WI-ID` with the actual identifier). A prompt containing only a file path is insufficient.
166
179
  ```bash
167
- $CREWX_CLI x "@{impl_agent} Execute Work Instruction WI-{ID}.
180
+ crewx x "@{impl_agent} Execute Work Instruction WI-{ID}.
168
181
 
169
182
  Full WI:
170
- {entire WI file content}
183
+ {paste the entire WI file content here}
171
184
 
172
185
  Report in the specified format when done." \
173
- --thread="WI-{ID}" --out=".crewx/result/WI-{ID}-impl.txt"
186
+ --thread="WI-{ID}" \
187
+ --out=".crewx/result/WI-{ID}-impl.txt"
188
+ ```
189
+ For `crewx x`, `--wait` requests synchronous execution; this differs from the `--wait=N` polling limit used by `crewx result` in the next step.
190
+ The prompt must contain the complete WI, not merely `docs/wi/WI-{ID}.md` or another file path.
191
+
192
+ 4. In agent context, the immediate task ID returned by `crewx x` is a receipt, not completion. Recover the terminal result from the foreground with:
193
+ ```bash
194
+ crewx result <task-id> --wait=N
174
195
  ```
196
+ Use `N <= 300`. Exit `124` means the task is still running, so repeat the result command using the next advised wait. Read both the terminal state (`success`, `failed`, `stopped`, or `runner_lost`) and the result body before deciding what happened.
197
+
198
+ If `--out` is used in receipt mode, the output file is written when the task completes, not when the receipt is returned. Do not wrap delegation with `timeout`, `&`, `nohup`, or any other background-process ownership mechanism.
175
199
 
176
- 5. Update WI status → `in-progress`
200
+ 5. After dispatch, update WI status → `in-progress`. Only after the terminal result is available and the ACs are verified may the WI be marked `done`.
177
201
 
178
- 6. Verify AC after completion report:
202
+ 6. Verify AC after the terminal completion report:
179
203
  ```bash
180
204
  npx tsc --noEmit
181
205
  npx vitest run src/server/ # or src/ui/
@@ -185,13 +209,9 @@ crewx skill wi issue WI-20260312-001
185
209
 
186
210
  ---
187
211
 
188
- ### list — List WIs
212
+ ### Manual procedure: list — List WIs
189
213
 
190
- ```bash
191
- crewx skill wi list
192
- crewx skill wi list --status=approved
193
- crewx skill wi list --status=in-progress
194
- ```
214
+ There is no implemented CLI route for `list`; use these repository commands manually.
195
215
 
196
216
  **Execution:**
197
217
  ```bash
@@ -204,13 +224,11 @@ grep -l "status: approved" docs/wi/WI-*.md 2>/dev/null
204
224
 
205
225
  ---
206
226
 
207
- ### review — Request lead review
227
+ ### Manual procedure: review — Request lead review
208
228
 
209
- Request a review from the `consultant` before issuing.
229
+ There is no implemented CLI route for `review`; request lead review manually before issuing.
210
230
 
211
- ```bash
212
- crewx skill wi review WI-20260312-001
213
- ```
231
+ Request a review from the `consultant` before issuing.
214
232
 
215
233
  ```bash
216
234
  $CREWX_CLI q "@{consultant} Review this Work Instruction.
@@ -271,6 +289,7 @@ When AC fails after agent completion report:
271
289
 
272
290
  ## References
273
291
 
274
- - Template: [docs/wi/WI-TEMPLATE.md](../../docs/wi/WI-TEMPLATE.md)
275
- - PM guide: [skills/pm/SKILL.md](../pm/SKILL.md)
276
- - Decomposition strategy: [docs/research/multi-agent-divide-conquer.md](../../docs/research/multi-agent-divide-conquer.md)
292
+ - Workspace WI template: `docs/wi/WI-TEMPLATE.md`
293
+ - Package-bundled seed: `packages/built-in/wi/templates/WI-TEMPLATE.md` (a separate file; these two templates are not claimed to be synchronized)
294
+ - PM guide: `skills/pm/SKILL.md`
295
+ - Decomposition strategy: `docs/research/multi-agent-divide-conquer.md`
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@crewx/wi",
3
- "version": "0.1.10-rc.115",
3
+ "version": "0.1.10-rc.117",
4
4
  "description": "Work Instruction engine for CrewX - structured task delegation",
5
5
  "type": "commonjs",
6
6
  "main": "dist/cli.js",
@@ -25,9 +25,9 @@
25
25
  "templates/"
26
26
  ],
27
27
  "dependencies": {
28
+ "@crewx/sdk": "0.9.0-rc.97",
28
29
  "@crewx/shared": "0.0.6",
29
- "@crewx/sdk": "0.9.0-rc.95",
30
- "@crewx/workflow": "0.3.22-rc.141"
30
+ "@crewx/workflow": "0.3.22-rc.143"
31
31
  },
32
32
  "peerDependencies": {
33
33
  "better-sqlite3": ">=12.0.0"