multi-agents-cli 1.1.54 → 1.1.55

package/core/templates/.agents/backend/API.md

@@ -193,6 +193,14 @@ Examples:
 This is mandatory. It is the only human-readable signal the user gets while
 the agent is working. Keep it specific and honest — not generic filler.
 
+### Output Mode
+
+Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
+
+- `full`     — no restrictions, behave normally
+- `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
+- `silent`   — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
+
 ---
 
 ## Workflow

package/core/templates/.agents/backend/AUTH.md

@@ -185,6 +185,14 @@ Examples:
 This is mandatory. It is the only human-readable signal the user gets while
 the agent is working. Keep it specific and honest — not generic filler.
 
+### Output Mode
+
+Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
+
+- `full`     — no restrictions, behave normally
+- `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
+- `silent`   — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
+
 ---
 
 ## Workflow

package/core/templates/.agents/backend/DB.md

@@ -192,6 +192,14 @@ Examples:
 This is mandatory. It is the only human-readable signal the user gets while
 the agent is working. Keep it specific and honest — not generic filler.
 
+### Output Mode
+
+Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
+
+- `full`     — no restrictions, behave normally
+- `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
+- `silent`   — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
+
 ---
 
 ## Workflow

package/core/templates/.agents/backend/EVENTS.md

@@ -191,6 +191,14 @@ Examples:
 This is mandatory. It is the only human-readable signal the user gets while
 the agent is working. Keep it specific and honest — not generic filler.
 
+### Output Mode
+
+Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
+
+- `full`     — no restrictions, behave normally
+- `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
+- `silent`   — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
+
 ---
 
 ## Workflow

package/core/templates/.agents/backend/INIT.md

@@ -175,6 +175,14 @@ Examples:
 This is mandatory. It is the only human-readable signal the user gets while
 the agent is working. Keep it specific and honest — not generic filler.
 
+### Output Mode
+
+Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
+
+- `full`     — no restrictions, behave normally
+- `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
+- `silent`   — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
+
 ---
 
 ## Workflow

package/core/templates/.agents/backend/JOBS.md

@@ -191,6 +191,14 @@ Examples:
 This is mandatory. It is the only human-readable signal the user gets while
 the agent is working. Keep it specific and honest — not generic filler.
 
+### Output Mode
+
+Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
+
+- `full`     — no restrictions, behave normally
+- `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
+- `silent`   — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
+
 ---
 
 ## Workflow

package/core/templates/.agents/backend/LOGIC.md

@@ -217,6 +217,14 @@ Examples:
 This is mandatory. It is the only human-readable signal the user gets while
 the agent is working. Keep it specific and honest — not generic filler.
 
+### Output Mode
+
+Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
+
+- `full`     — no restrictions, behave normally
+- `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
+- `silent`   — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
+
 ---
 
 ## Workflow

package/core/templates/.agents/backend/TESTING.md

@@ -199,6 +199,14 @@ Examples:
 This is mandatory. It is the only human-readable signal the user gets while
 the agent is working. Keep it specific and honest — not generic filler.
 
+### Output Mode
+
+Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
+
+- `full`     — no restrictions, behave normally
+- `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
+- `silent`   — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
+
 ---
 
 ## Workflow

package/core/templates/.agents/client/ACCESSIBILITY.md

@@ -194,6 +194,14 @@ Examples:
 This is mandatory. It is the only human-readable signal the user gets while
 the agent is working. Keep it specific and honest — not generic filler.
 
+### Output Mode
+
+Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
+
+- `full`     — no restrictions, behave normally
+- `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
+- `silent`   — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
+
 ---
 
 ## Workflow

package/core/templates/.agents/client/FORMS.md

@@ -179,6 +179,14 @@ Examples:
 This is mandatory. It is the only human-readable signal the user gets while
 the agent is working. Keep it specific and honest — not generic filler.
 
+### Output Mode
+
+Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
+
+- `full`     — no restrictions, behave normally
+- `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
+- `silent`   — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
+
 ---
 
 ## Workflow

package/core/templates/.agents/client/LOGIC.md

@@ -211,6 +211,14 @@ Examples:
 This is mandatory. It is the only human-readable signal the user gets while
 the agent is working. Keep it specific and honest — not generic filler.
 
+### Output Mode
+
+Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
+
+- `full`     — no restrictions, behave normally
+- `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
+- `silent`   — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
+
 ---
 
 ## Workflow

package/core/templates/.agents/client/ROUTING.md

@@ -180,6 +180,14 @@ Examples:
 This is mandatory. It is the only human-readable signal the user gets while
 the agent is working. Keep it specific and honest — not generic filler.
 
+### Output Mode
+
+Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
+
+- `full`     — no restrictions, behave normally
+- `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
+- `silent`   — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
+
 ---
 
 ## Workflow

package/core/templates/.agents/client/TESTING.md

@@ -181,6 +181,14 @@ Examples:
 This is mandatory. It is the only human-readable signal the user gets while
 the agent is working. Keep it specific and honest — not generic filler.
 
+### Output Mode
+
+Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
+
+- `full`     — no restrictions, behave normally
+- `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
+- `silent`   — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
+
 ---
 
 ## Workflow

package/core/templates/.agents/client/UI.md

@@ -172,6 +172,14 @@ Examples:
 This is mandatory. It is the only human-readable signal the user gets while
 the agent is working. Keep it specific and honest — not generic filler.
 
+### Output Mode
+
+Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
+
+- `full`     — no restrictions, behave normally
+- `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
+- `silent`   — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
+
 ---
 
 ## Workflow

package/core/templates/.agents/shared/SECURITY.md

@@ -194,6 +194,14 @@ Examples:
 This is mandatory. It is the only human-readable signal the user gets while
 the agent is working. Keep it specific and honest — not generic filler.
 
+### Output Mode
+
+Read `output_mode` from `.claude-scope` at session start. Honor it for the entire session:
+
+- `full`     — no restrictions, behave normally
+- `insights` — emit `▶ narration` lines before each phase; suppress routine file read/write commentary in between
+- `silent`   — emit only git operations, errors, and verification results; show `✢ working...` between phases, no other output
+
 ---
 
 ## Workflow

package/core/templates/CLAUDE.md

@@ -146,7 +146,12 @@ the agent must:
    # then take the parent directory of that path - this is PROJECT_ROOT
    ```
    Store this as PROJECT_ROOT. All subsequent paths (`.scaffold/`, `BUILD_STATE.md`, `CONTRACTS.md`, etc.) resolve from PROJECT_ROOT, never from the worktree directory.
-2. Read `BUILD_STATE.md` at PROJECT_ROOT - understand what has been built
+2. Read `output_mode` from `.claude-scope` in the current worktree directory:
+   - `full`     — default, no restrictions on output
+   - `insights` — emit `▶ narration` lines before each phase only; suppress routine file read/write commentary
+   - `silent`   — emit only git operations, errors, and verification results; show `✢ working...` between phases
+   Store this as OUTPUT_MODE. Apply it consistently for the entire session.
+3. Read `BUILD_STATE.md` at PROJECT_ROOT - understand what has been built
 3. Check if `TASK.md` exists in the current directory
 4. If yes - read it and verify dependencies are met against BUILD_STATE.md
 5. If dependencies not met - surface what is missing and propose options

package/core/workflow/agent.js

@@ -486,6 +486,7 @@ project      : ${project}
 agent        : ${agent}
 branch       : ${branchName}
 worktree     : ${worktreePath}
+output_mode  : ${outputMode}
 
 ## Scope Verification Rule
 Before doing anything else, verify:
@@ -1733,6 +1734,17 @@ ${excludedUrls}
   );
   const skipPermissions = permIdx === 0;
 
+  const outputModeIdx = await arrowSelect(
+    'Agent output mode for this session?',
+    [
+      { label: `${green('→')} Full output      ${dim('— see everything (default)')}` },
+      { label: `${green('→')} Insights only    ${dim('— phase narrations + important blocks')}` },
+      { label: `${green('→')} Silent           ${dim('— spinner + critical blocks only')}` },
+    ],
+    rl
+  );
+  const outputMode = ['full', 'insights', 'silent'][outputModeIdx];
+
   sessionLoop: while (true) {
   const sessionIdx = await arrowSelect('How would you like to start the session?', [
     { label: `${green('→')} IDE + new terminal ${dim('(Claude Code CLI)')}  ${dim('← recommended')}` },

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "multi-agents-cli",
-  "version": "1.1.54",
+  "version": "1.1.55",
   "description": "Multi-agent workflow orchestration for Claude Code — isolated git worktrees, structured state tracking, autonomous task chaining",
   "keywords": [
     "claude-code",