multi-agents-cli 1.1.53 → 1.1.55

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.
@@ -177,6 +177,32 @@ These apply to every API task regardless of framework.
177
177
 
178
178
  ---
179
179
 
180
+ ## Progress Narration
181
+
182
+ Before starting each major build phase, emit one plain-English status line:
183
+
184
+ ```
185
+ ▶ [Phase name] — [what is being built and why, one line]
186
+ ```
187
+
188
+ Examples:
189
+ - `▶ Exploring existing components — checking current patterns before writing anything`
190
+ - `▶ Building UI primitives — Button, Badge, Card as presentational shells`
191
+ - `▶ Validating output — running tsc and dev server to confirm zero errors`
192
+
193
+ This is mandatory. It is the only human-readable signal the user gets while
194
+ the agent is working. Keep it specific and honest — not generic filler.
195
+
196
+ ### Output Mode
197
+
198
+ Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
199
+
200
+ - `full` — no restrictions, behave normally
201
+ - `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
202
+ - `silent` — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
203
+
204
+ ---
205
+
180
206
  ## Workflow
181
207
 
182
208
  ```
@@ -169,6 +169,32 @@ These apply to every auth task regardless of framework.
169
169
 
170
170
  ---
171
171
 
172
+ ## Progress Narration
173
+
174
+ Before starting each major build phase, emit one plain-English status line:
175
+
176
+ ```
177
+ ▶ [Phase name] — [what is being built and why, one line]
178
+ ```
179
+
180
+ Examples:
181
+ - `▶ Exploring existing components — checking current patterns before writing anything`
182
+ - `▶ Building UI primitives — Button, Badge, Card as presentational shells`
183
+ - `▶ Validating output — running tsc and dev server to confirm zero errors`
184
+
185
+ This is mandatory. It is the only human-readable signal the user gets while
186
+ the agent is working. Keep it specific and honest — not generic filler.
187
+
188
+ ### Output Mode
189
+
190
+ Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
191
+
192
+ - `full` — no restrictions, behave normally
193
+ - `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
194
+ - `silent` — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
195
+
196
+ ---
197
+
172
198
  ## Workflow
173
199
 
174
200
  ```
@@ -176,6 +176,32 @@ These apply to every DB task regardless of framework.
176
176
 
177
177
  ---
178
178
 
179
+ ## Progress Narration
180
+
181
+ Before starting each major build phase, emit one plain-English status line:
182
+
183
+ ```
184
+ ▶ [Phase name] — [what is being built and why, one line]
185
+ ```
186
+
187
+ Examples:
188
+ - `▶ Exploring existing components — checking current patterns before writing anything`
189
+ - `▶ Building UI primitives — Button, Badge, Card as presentational shells`
190
+ - `▶ Validating output — running tsc and dev server to confirm zero errors`
191
+
192
+ This is mandatory. It is the only human-readable signal the user gets while
193
+ the agent is working. Keep it specific and honest — not generic filler.
194
+
195
+ ### Output Mode
196
+
197
+ Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
198
+
199
+ - `full` — no restrictions, behave normally
200
+ - `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
201
+ - `silent` — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
202
+
203
+ ---
204
+
179
205
  ## Workflow
180
206
 
181
207
  ```
@@ -175,6 +175,32 @@ These apply to every events task regardless of framework.
175
175
 
176
176
  ---
177
177
 
178
+ ## Progress Narration
179
+
180
+ Before starting each major build phase, emit one plain-English status line:
181
+
182
+ ```
183
+ ▶ [Phase name] — [what is being built and why, one line]
184
+ ```
185
+
186
+ Examples:
187
+ - `▶ Exploring existing components — checking current patterns before writing anything`
188
+ - `▶ Building UI primitives — Button, Badge, Card as presentational shells`
189
+ - `▶ Validating output — running tsc and dev server to confirm zero errors`
190
+
191
+ This is mandatory. It is the only human-readable signal the user gets while
192
+ the agent is working. Keep it specific and honest — not generic filler.
193
+
194
+ ### Output Mode
195
+
196
+ Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
197
+
198
+ - `full` — no restrictions, behave normally
199
+ - `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
200
+ - `silent` — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
201
+
202
+ ---
203
+
178
204
  ## Workflow
179
205
 
180
206
  ```
@@ -159,6 +159,32 @@ is responsible for populating the actual values in `.env.*` files.
159
159
 
160
160
  ---
161
161
 
162
+ ## Progress Narration
163
+
164
+ Before starting each major build phase, emit one plain-English status line:
165
+
166
+ ```
167
+ ▶ [Phase name] — [what is being built and why, one line]
168
+ ```
169
+
170
+ Examples:
171
+ - `▶ Exploring existing components — checking current patterns before writing anything`
172
+ - `▶ Building UI primitives — Button, Badge, Card as presentational shells`
173
+ - `▶ Validating output — running tsc and dev server to confirm zero errors`
174
+
175
+ This is mandatory. It is the only human-readable signal the user gets while
176
+ the agent is working. Keep it specific and honest — not generic filler.
177
+
178
+ ### Output Mode
179
+
180
+ Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
181
+
182
+ - `full` — no restrictions, behave normally
183
+ - `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
184
+ - `silent` — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
185
+
186
+ ---
187
+
162
188
  ## Workflow
163
189
 
164
190
  ```
@@ -175,6 +175,32 @@ These apply to every jobs task regardless of framework.
175
175
 
176
176
  ---
177
177
 
178
+ ## Progress Narration
179
+
180
+ Before starting each major build phase, emit one plain-English status line:
181
+
182
+ ```
183
+ ▶ [Phase name] — [what is being built and why, one line]
184
+ ```
185
+
186
+ Examples:
187
+ - `▶ Exploring existing components — checking current patterns before writing anything`
188
+ - `▶ Building UI primitives — Button, Badge, Card as presentational shells`
189
+ - `▶ Validating output — running tsc and dev server to confirm zero errors`
190
+
191
+ This is mandatory. It is the only human-readable signal the user gets while
192
+ the agent is working. Keep it specific and honest — not generic filler.
193
+
194
+ ### Output Mode
195
+
196
+ Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
197
+
198
+ - `full` — no restrictions, behave normally
199
+ - `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
200
+ - `silent` — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
201
+
202
+ ---
203
+
178
204
  ## Workflow
179
205
 
180
206
  ```
@@ -201,6 +201,32 @@ These apply to every backend logic task regardless of framework.
201
201
 
202
202
  ---
203
203
 
204
+ ## Progress Narration
205
+
206
+ Before starting each major build phase, emit one plain-English status line:
207
+
208
+ ```
209
+ ▶ [Phase name] — [what is being built and why, one line]
210
+ ```
211
+
212
+ Examples:
213
+ - `▶ Exploring existing components — checking current patterns before writing anything`
214
+ - `▶ Building UI primitives — Button, Badge, Card as presentational shells`
215
+ - `▶ Validating output — running tsc and dev server to confirm zero errors`
216
+
217
+ This is mandatory. It is the only human-readable signal the user gets while
218
+ the agent is working. Keep it specific and honest — not generic filler.
219
+
220
+ ### Output Mode
221
+
222
+ Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
223
+
224
+ - `full` — no restrictions, behave normally
225
+ - `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
226
+ - `silent` — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
227
+
228
+ ---
229
+
204
230
  ## Workflow
205
231
 
206
232
  ```
@@ -183,6 +183,32 @@ These apply to every backend testing task regardless of framework.
183
183
 
184
184
  ---
185
185
 
186
+ ## Progress Narration
187
+
188
+ Before starting each major build phase, emit one plain-English status line:
189
+
190
+ ```
191
+ ▶ [Phase name] — [what is being built and why, one line]
192
+ ```
193
+
194
+ Examples:
195
+ - `▶ Exploring existing components — checking current patterns before writing anything`
196
+ - `▶ Building UI primitives — Button, Badge, Card as presentational shells`
197
+ - `▶ Validating output — running tsc and dev server to confirm zero errors`
198
+
199
+ This is mandatory. It is the only human-readable signal the user gets while
200
+ the agent is working. Keep it specific and honest — not generic filler.
201
+
202
+ ### Output Mode
203
+
204
+ Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
205
+
206
+ - `full` — no restrictions, behave normally
207
+ - `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
208
+ - `silent` — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
209
+
210
+ ---
211
+
186
212
  ## Workflow
187
213
 
188
214
  ```
@@ -178,6 +178,32 @@ These apply to every accessibility task regardless of framework.
178
178
 
179
179
  ---
180
180
 
181
+ ## Progress Narration
182
+
183
+ Before starting each major build phase, emit one plain-English status line:
184
+
185
+ ```
186
+ ▶ [Phase name] — [what is being built and why, one line]
187
+ ```
188
+
189
+ Examples:
190
+ - `▶ Exploring existing components — checking current patterns before writing anything`
191
+ - `▶ Building UI primitives — Button, Badge, Card as presentational shells`
192
+ - `▶ Validating output — running tsc and dev server to confirm zero errors`
193
+
194
+ This is mandatory. It is the only human-readable signal the user gets while
195
+ the agent is working. Keep it specific and honest — not generic filler.
196
+
197
+ ### Output Mode
198
+
199
+ Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
200
+
201
+ - `full` — no restrictions, behave normally
202
+ - `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
203
+ - `silent` — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
204
+
205
+ ---
206
+
181
207
  ## Workflow
182
208
 
183
209
  ```
@@ -163,6 +163,32 @@ These apply to every form task regardless of framework.
163
163
 
164
164
  ---
165
165
 
166
+ ## Progress Narration
167
+
168
+ Before starting each major build phase, emit one plain-English status line:
169
+
170
+ ```
171
+ ▶ [Phase name] — [what is being built and why, one line]
172
+ ```
173
+
174
+ Examples:
175
+ - `▶ Exploring existing components — checking current patterns before writing anything`
176
+ - `▶ Building UI primitives — Button, Badge, Card as presentational shells`
177
+ - `▶ Validating output — running tsc and dev server to confirm zero errors`
178
+
179
+ This is mandatory. It is the only human-readable signal the user gets while
180
+ the agent is working. Keep it specific and honest — not generic filler.
181
+
182
+ ### Output Mode
183
+
184
+ Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
185
+
186
+ - `full` — no restrictions, behave normally
187
+ - `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
188
+ - `silent` — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
189
+
190
+ ---
191
+
166
192
  ## Workflow
167
193
 
168
194
  ```
@@ -195,6 +195,32 @@ These apply to every client logic task regardless of framework.
195
195
 
196
196
  ---
197
197
 
198
+ ## Progress Narration
199
+
200
+ Before starting each major build phase, emit one plain-English status line:
201
+
202
+ ```
203
+ ▶ [Phase name] — [what is being built and why, one line]
204
+ ```
205
+
206
+ Examples:
207
+ - `▶ Exploring existing components — checking current patterns before writing anything`
208
+ - `▶ Building UI primitives — Button, Badge, Card as presentational shells`
209
+ - `▶ Validating output — running tsc and dev server to confirm zero errors`
210
+
211
+ This is mandatory. It is the only human-readable signal the user gets while
212
+ the agent is working. Keep it specific and honest — not generic filler.
213
+
214
+ ### Output Mode
215
+
216
+ Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
217
+
218
+ - `full` — no restrictions, behave normally
219
+ - `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
220
+ - `silent` — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
221
+
222
+ ---
223
+
198
224
  ## Workflow
199
225
 
200
226
  ```
@@ -164,6 +164,32 @@ These apply to every routing task regardless of framework.
164
164
 
165
165
  ---
166
166
 
167
+ ## Progress Narration
168
+
169
+ Before starting each major build phase, emit one plain-English status line:
170
+
171
+ ```
172
+ ▶ [Phase name] — [what is being built and why, one line]
173
+ ```
174
+
175
+ Examples:
176
+ - `▶ Exploring existing components — checking current patterns before writing anything`
177
+ - `▶ Building UI primitives — Button, Badge, Card as presentational shells`
178
+ - `▶ Validating output — running tsc and dev server to confirm zero errors`
179
+
180
+ This is mandatory. It is the only human-readable signal the user gets while
181
+ the agent is working. Keep it specific and honest — not generic filler.
182
+
183
+ ### Output Mode
184
+
185
+ Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
186
+
187
+ - `full` — no restrictions, behave normally
188
+ - `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
189
+ - `silent` — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
190
+
191
+ ---
192
+
167
193
  ## Workflow
168
194
 
169
195
  ```
@@ -165,6 +165,32 @@ These apply to every testing task regardless of framework.
165
165
 
166
166
  ---
167
167
 
168
+ ## Progress Narration
169
+
170
+ Before starting each major build phase, emit one plain-English status line:
171
+
172
+ ```
173
+ ▶ [Phase name] — [what is being built and why, one line]
174
+ ```
175
+
176
+ Examples:
177
+ - `▶ Exploring existing components — checking current patterns before writing anything`
178
+ - `▶ Building UI primitives — Button, Badge, Card as presentational shells`
179
+ - `▶ Validating output — running tsc and dev server to confirm zero errors`
180
+
181
+ This is mandatory. It is the only human-readable signal the user gets while
182
+ the agent is working. Keep it specific and honest — not generic filler.
183
+
184
+ ### Output Mode
185
+
186
+ Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
187
+
188
+ - `full` — no restrictions, behave normally
189
+ - `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
190
+ - `silent` — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
191
+
192
+ ---
193
+
168
194
  ## Workflow
169
195
 
170
196
  ```
@@ -156,6 +156,32 @@ These apply to every UI task regardless of framework.
156
156
 
157
157
  ---
158
158
 
159
+ ## Progress Narration
160
+
161
+ Before starting each major build phase, emit one plain-English status line:
162
+
163
+ ```
164
+ ▶ [Phase name] — [what is being built and why, one line]
165
+ ```
166
+
167
+ Examples:
168
+ - `▶ Exploring existing components — checking current patterns before writing anything`
169
+ - `▶ Building UI primitives — Button, Badge, Card as presentational shells`
170
+ - `▶ Validating output — running tsc and dev server to confirm zero errors`
171
+
172
+ This is mandatory. It is the only human-readable signal the user gets while
173
+ the agent is working. Keep it specific and honest — not generic filler.
174
+
175
+ ### Output Mode
176
+
177
+ Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
178
+
179
+ - `full` — no restrictions, behave normally
180
+ - `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
181
+ - `silent` — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
182
+
183
+ ---
184
+
159
185
  ## Workflow
160
186
 
161
187
  ```
@@ -178,6 +178,32 @@ These apply to every security task regardless of framework or project type.
178
178
 
179
179
  ---
180
180
 
181
+ ## Progress Narration
182
+
183
+ Before starting each major build phase, emit one plain-English status line:
184
+
185
+ ```
186
+ ▶ [Phase name] — [what is being built and why, one line]
187
+ ```
188
+
189
+ Examples:
190
+ - `▶ Exploring existing components — checking current patterns before writing anything`
191
+ - `▶ Building UI primitives — Button, Badge, Card as presentational shells`
192
+ - `▶ Validating output — running tsc and dev server to confirm zero errors`
193
+
194
+ This is mandatory. It is the only human-readable signal the user gets while
195
+ the agent is working. Keep it specific and honest — not generic filler.
196
+
197
+ ### Output Mode
198
+
199
+ Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
200
+
201
+ - `full` — no restrictions, behave normally
202
+ - `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
203
+ - `silent` — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
204
+
205
+ ---
206
+
181
207
  ## Workflow
182
208
 
183
209
  ```
@@ -146,7 +146,12 @@ the agent must:
146
146
  # then take the parent directory of that path - this is PROJECT_ROOT
147
147
  ```
148
148
  Store this as PROJECT_ROOT. All subsequent paths (`.scaffold/`, `BUILD_STATE.md`, `CONTRACTS.md`, etc.) resolve from PROJECT_ROOT, never from the worktree directory.
149
- 2. Read `BUILD_STATE.md` at PROJECT_ROOT - understand what has been built
149
+ 2. Read `output_mode` from `.claude-scope` in the current worktree directory:
150
+ - `full` — default, no restrictions on output
151
+ - `insights` — emit `▶ narration` lines before each phase only; suppress routine file read/write commentary
152
+ - `silent` — emit only git operations, errors, and verification results; show `✢ working...` between phases
153
+ Store this as OUTPUT_MODE. Apply it consistently for the entire session.
154
+ 3. Read `BUILD_STATE.md` at PROJECT_ROOT - understand what has been built
150
155
  3. Check if `TASK.md` exists in the current directory
151
156
  4. If yes - read it and verify dependencies are met against BUILD_STATE.md
152
157
  5. If dependencies not met - surface what is missing and propose options
@@ -486,6 +486,7 @@ project : ${project}
486
486
  agent : ${agent}
487
487
  branch : ${branchName}
488
488
  worktree : ${worktreePath}
489
+ output_mode : ${outputMode}
489
490
 
490
491
  ## Scope Verification Rule
491
492
  Before doing anything else, verify:
@@ -1733,6 +1734,17 @@ ${excludedUrls}
1733
1734
  );
1734
1735
  const skipPermissions = permIdx === 0;
1735
1736
 
1737
+ const outputModeIdx = await arrowSelect(
1738
+ 'Agent output mode for this session?',
1739
+ [
1740
+ { label: `${green('→')} Full output ${dim('— see everything (default)')}` },
1741
+ { label: `${green('→')} Insights only ${dim('— phase narrations + important blocks')}` },
1742
+ { label: `${green('→')} Silent ${dim('— spinner + critical blocks only')}` },
1743
+ ],
1744
+ rl
1745
+ );
1746
+ const outputMode = ['full', 'insights', 'silent'][outputModeIdx];
1747
+
1736
1748
  sessionLoop: while (true) {
1737
1749
  const sessionIdx = await arrowSelect('How would you like to start the session?', [
1738
1750
  { label: `${green('→')} IDE + new terminal ${dim('(Claude Code CLI)')} ${dim('← recommended')}` },
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "multi-agents-cli",
3
- "version": "1.1.53",
3
+ "version": "1.1.55",
4
4
  "description": "Multi-agent workflow orchestration for Claude Code — isolated git worktrees, structured state tracking, autonomous task chaining",
5
5
  "keywords": [
6
6
  "claude-code",