ProjectAI00 · Copilot · Mar 4, 2026 · Mar 4, 2026
diff --git a/README.md b/README.md
@@ -79,7 +79,8 @@ imi goal "name" --why "why now"       # Create a goal
 imi task <goal_id> "title" --why "..." # Add a task
 imi log "direction note"              # Log strategic direction
 imi decide "what" "why"               # Log a decision
-imi orchestrate <goal_id> --workers 8 -- <cmd ...>  # Parallel worker loop
+imi orchestrate <goal_id> --workers 8 -- <cmd ...>  # Parallel worker loop (hankweave)
+imi orchestrate <goal_id> --workers 4 --cli auto    # Auto-detect agent CLI from environment
 ```
 
 ## The Loop
@@ -151,4 +152,21 @@ imi next --agent engineer-b --toon   # Agent B claims task 2 (different task)
 imi next --agent engineer-c --toon   # Agent C claims task 3
 ```
 
+Or spin up a parallel worker pool that auto-selects the right CLI for the current environment:
+
+```bash
+imi orchestrate <goal_id> --workers 4 --cli auto
+```
+
+`--cli auto` detects the running agent environment and routes accordingly:
+
+| Environment variable | Detected agent |
+|---|---|
+| `CLAUDE_CODE_SSE_PORT` or `CLAUDE_CODE_ENTRYPOINT` | Claude Code |
+| `OPENCODE_SESSION` | OpenCode |
+| `GH_COPILOT_SESSION_ID` or `COPILOT_AGENT_SESSION` | GitHub Copilot |
+| _(none)_ | hankweave (default) |
+
+You can also specify the CLI explicitly: `--cli claude`, `--cli opencode`, `--cli codex`, or `--cli hankweave`.
+
 If a task is abandoned, IMI auto-releases it after 30 minutes. The next agent picks it up with full failure context.
diff --git a/skills/imi/SKILL.md b/skills/imi/SKILL.md
@@ -0,0 +1,87 @@
+---
+name: imi
+description: >
+  Activate when: the project has a .imi/ directory, or the user asks about
+  what to build, what's next, tasks, goals, decisions, progress, or where we
+  left off. Always run `imi context` before responding. Install if missing: bunx imi-agent
+license: MIT
+metadata:
+  author: ProjectAI00
+  version: "3.0"
+compatibility: Requires imi CLI. Install: bunx imi-agent
+allowed-tools: Bash(imi:*)
+---
+
+# IMI — Agent Instruction Manual
+
+IMI is the PM thinking layer between humans and agents. It keeps human intent, direction, and decisions persistent so every session stays aligned without re-briefing. Goals, tasks, decisions, and direction notes live in `.imi/state.db`.
+
+**The 3 questions IMI must always answer:**
+1. What are we building?
+2. How is it going?
+3. Are we still aligned with intent?
+
+**Authority order:** Direction notes + decisions → goals → tasks → execution. If work cannot trace to a direction or decision, flag it before doing anything.
+
+---
+
+## ⛔ HARD STOP — before you do anything
+
+DO NOT:
+- `cat`, `grep`, `ls`, or `sqlite3` any file inside `.imi/`
+- Use session memory, built-in todos, or conversation history as project state
+- Answer any question about project status without first running `imi context`
+- Create a goal or task without filling in `why` and `success_signal`
+- Execute work that cannot be traced to a goal in the DB
+
+---
+
+## Every session — no exceptions
+
+```bash
+imi context
+```
+
+Run this before your first response. Every session. No exceptions. Then ask: does the user's request map to a goal in the DB? If you can't point to one, say so before doing anything.
+
+---
+
+## Mode routing
+
+Detect which mode applies and load the full instructions:
+
+| Situation | Mode | Load |
+|---|---|---|
+| User is checking status, making decisions, thinking out loud | **Ops** | `cat ~/.copilot/skills/imi/ops-mode.md` |
+| User wants to plan goals, decompose work, write task specs | **Plan** | `cat ~/.copilot/skills/imi/plan-mode.md` |
+| User wants to execute a task, do the work | **Execute** | `cat ~/.copilot/skills/imi/execute-mode.md` |
+| Writing `imi complete`, `imi fail`, or memory entries | **Voice** | `cat ~/.copilot/skills/imi/ai-voice.md` |
+
+Load the relevant mode file before proceeding. It contains the full behavioral contract for that mode.
+
+---
+
+## Quick command reference
+
+```bash
+imi context                        # start every session here
+imi think                          # is this still the right thing to build?
+imi plan                           # full goal + task list
+imi decide "what" "why"            # log a firm decision
+imi log "note"                     # log a direction, instinct, or observation
+imi goal "<name>" "<desc>" <pri> "<why>" "<for_who>" "<success_signal>"
+imi task <goal_id> "<title>" --why "<reason>" --acceptance-criteria "<done looks like>" --relevant-files "<files>"
+imi complete <task_id> "summary"   # mark done — never skip this
+imi mlesson "what went wrong"      # store a lesson after a corrected mistake
+imi check                          # verification state
+```
+
+---
+
+## If imi is not installed
+
+```bash
+bunx imi-agent
+```
+
+Then re-run `imi context`.
diff --git a/tests/integration.sh b/tests/integration.sh
@@ -375,6 +375,71 @@ else
   fail "orchestrate: completion summaries missing" "DB said: $DB_OUT"
 fi
 
+# ─────────────────────────────────────────────────────────────
+# 9D. ORCHESTRATE --cli auto (auto-selection)
+# ─────────────────────────────────────────────────────────────
+echo ""
+echo "── 9D. Orchestrate --cli auto ──────────────────────────"
+
+run add-goal "Auto-select goal" "Verify --cli auto selects correct agent CLI"
+AUTO_GOAL_ID=$(echo "$CMD_OUT" | grep -oE '[a-z0-9]{14,}' | head -1)
+if [[ -z "$AUTO_GOAL_ID" ]]; then
+  db_query "SELECT id FROM goals ORDER BY rowid DESC LIMIT 1;"
+  AUTO_GOAL_ID="$DB_OUT"
+fi
+
+run add-task "$AUTO_GOAL_ID" "auto-select task 1" "auto worker one"
+run add-task "$AUTO_GOAL_ID" "auto-select task 2" "auto worker two"
+
+# Create a fake 'claude' binary that simply exits 0 (simulates a real agent run).
+# $FAKE_BIN_DIR lives inside $TEST_DIR which is removed by the EXIT trap above.
+FAKE_BIN_DIR="$TEST_DIR/fake-bin"
+mkdir -p "$FAKE_BIN_DIR"
+cat > "$FAKE_BIN_DIR/claude" <<'EOF'
+#!/usr/bin/env bash
+exit 0
+EOF
+chmod +x "$FAKE_BIN_DIR/claude"
+
+# With CLAUDE_CODE_ENTRYPOINT set, --cli auto should select the 'claude' path.
+# The resolved command becomes: sh -c 'claude -p "$(cat "$IMI_TASK_CONTEXT_FILE")" ...'
+# Our fake claude ignores arguments and exits 0, so wrap auto-completes the task.
+CLAUDE_CODE_ENTRYPOINT=1 PATH="$FAKE_BIN_DIR:$PATH" \
+  "$IMI_BIN" orchestrate "$AUTO_GOAL_ID" \
+    --workers 2 --max-tasks 2 --cli auto --agent-prefix auto-bench --ping-secs 1 --checkpoint-secs 1 \
+  2>&1 && CMD_EXIT=0 || CMD_EXIT=$?
+if [[ "$CMD_EXIT" -eq 0 ]]; then
+  pass "orchestrate --cli auto (CLAUDE_CODE_ENTRYPOINT) exits 0"
+else
+  fail "orchestrate --cli auto (CLAUDE_CODE_ENTRYPOINT) exited $CMD_EXIT"
+fi
+
+db_query "SELECT COUNT(*) FROM tasks WHERE goal_id='$AUTO_GOAL_ID' AND status='done';"
+if [[ "$DB_OUT" == "2" ]]; then
+  pass "orchestrate --cli auto: all tasks completed"
+else
+  fail "orchestrate --cli auto: expected 2 done tasks" "DB said: $DB_OUT"
+fi
+
+# With no agent env vars, --cli auto should fall back to hankweave (imi run).
+# Use --max-tasks 0 so no workers are launched — just verifies the flag is accepted.
+run add-goal "Auto-select fallback" "Verify --cli auto with no env vars"
+AUTO_FB_GOAL_ID=$(echo "$CMD_OUT" | grep -oE '[a-z0-9]{14,}' | head -1)
+if [[ -z "$AUTO_FB_GOAL_ID" ]]; then
+  db_query "SELECT id FROM goals ORDER BY rowid DESC LIMIT 1;"
+  AUTO_FB_GOAL_ID="$DB_OUT"
+fi
+(
+  # Unset all env vars documented in README's auto-detection table.
+  unset CLAUDE_CODE_SSE_PORT CLAUDE_CODE_ENTRYPOINT OPENCODE_SESSION GH_COPILOT_SESSION_ID COPILOT_AGENT_SESSION
+  "$IMI_BIN" orchestrate "$AUTO_FB_GOAL_ID" --workers 1 --max-tasks 0 --cli auto
+) && CMD_EXIT=0 || CMD_EXIT=$?
+if [[ "$CMD_EXIT" -eq 0 ]]; then
+  pass "orchestrate --cli auto (no env vars fallback) exits 0"
+else
+  fail "orchestrate --cli auto (no env vars fallback) exited $CMD_EXIT"
+fi
+
 # ═════════════════════════════════════════════════════════════
 # 10. MEMORY ADD + LIST
 # ═════════════════════════════════════════════════════════════