aether-colony 5.0.0 → 5.1.0

package/.aether/commands/resume.yaml

@@ -0,0 +1,398 @@
+---
+name: ant:resume
+description: "Resume Previous Session"
+body: |
+  # /ant:resume — Resume Previous Session
+
+  Resume work after `/clear` or in a new session. Reads colony state, detects codebase drift, and gives you a clear "do this next" recommendation.
+
+  ## Usage
+
+  ```bash
+  /ant:resume
+  ```
+
+  ---
+
+  ## Implementation
+
+  Execute the following steps in order when the user runs `/ant:resume`.
+
+  ---
+
+  ### Step 1: Read Session State
+
+  {{TOOL_PREFIX "Restoring colony session..."}}:
+  ```bash
+  bash .aether/aether-utils.sh session-read
+  ```
+
+  Parse the JSON result.
+
+  - If `exists` is `false`: display the following and **stop**:
+
+  ```
+  {{#claude}}━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━{{/claude}}
+  {{#opencode}}═══════════════════════════════════════════════════{{/opencode}}
+  RESUME SESSION
+
+  No previous session found.
+
+  Start fresh: /ant:init "your goal"
+  Or check: /ant:status
+  {{#claude}}━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━{{/claude}}
+  {{#opencode}}═══════════════════════════════════════════════════{{/opencode}}
+  ```
+
+  - If `exists` is `true`: extract from the session data:
+    - `colony_goal`
+    - `current_phase`
+    - `last_command`
+    - `suggested_next`
+    - `baseline_commit`
+    - `session_id`
+
+  ---
+
+  ### Step 2: Read COLONY_STATE.json (Authoritative Source)
+
+  Use the Read tool to read `.aether/data/COLONY_STATE.json`.
+
+  COLONY_STATE.json is the authoritative source for goal and state (session.json may be stale). Extract:
+  - `goal` (use this as authoritative, overriding session.json colony_goal)
+  - `state` (READY, PLANNING, EXECUTING, PAUSED)
+  - `current_phase`
+  - `plan.phases` (array with id, name, status for each phase)
+  - `plan.generated_at`
+  - `memory.decisions` (flat list — do NOT distinguish user vs Claude origin)
+  - `events` (last 5 for recent activity context)
+
+  If the file is missing or the JSON cannot be parsed, **stop immediately** and display:
+
+  ```
+  State file missing or corrupted.
+
+  Options:
+  1. Start fresh with /ant:init "goal"
+  2. Try to recover (I'll look for backup files)
+
+  What would you like to do?
+  ```
+
+  Do NOT proceed with stale or fabricated data.
+
+  ---
+
+  ### Step 3: Read Pheromone Signals
+
+  {{#claude}}
+  Use the Read tool to read `.aether/data/constraints.json`.
+
+  Extract the following top-level keys:
+  - `focus` array — active focus signals (if key missing, treat as empty array)
+  - `constraints` array — active redirect/constraint signals (if key missing, treat as empty array)
+
+  If the file is missing: skip silently (no pheromones active).
+
+  Pheromones persist until explicitly cleared — no decay.
+  {{/claude}}
+  {{#opencode}}
+  {{TOOL_PREFIX "Loading active pheromone signals..."}}:
+  ```bash
+  bash .aether/aether-utils.sh pheromone-read all
+  ```
+
+  Parse the JSON result. Extract `.result.signals` array.
+
+  - If `ok` is `true` and `.result.signals` is non-empty: store signals for dashboard rendering in Step 8
+  - If `ok` is `true` and `.result.signals` is empty: no active pheromones (skip in dashboard)
+  - If the command fails or returns an error: skip silently (no pheromones active)
+
+  Note: pheromone-read applies decay calculation automatically. The `effective_strength` field reflects current signal strength after time-based decay. Signals below 0.1 effective strength are already filtered out.
+  {{/opencode}}
+
+  ---
+
+  ### Step 4: Read CONTEXT.md
+
+  Use the Read tool to read `.aether/CONTEXT.md` if it exists.
+
+  If missing: fall back to COLONY_STATE.json for narrative context. Note: "Context document not found — reconstructing from state."
+
+  ---
+
+  ### Step 5: Drift Detection
+
+  Extract `baseline_commit` from the session.json data read in Step 1.
+
+  ```bash
+  current_commit=$(git rev-parse HEAD 2>/dev/null || echo "")
+  ```
+
+  If `baseline_commit` is non-empty and differs from `current_commit`:
+
+  ```bash
+  commit_count=$(git rev-list --count "$baseline_commit..HEAD" 2>/dev/null || echo "0")
+  changed_count=$(git diff --stat "$baseline_commit" HEAD 2>/dev/null | tail -1 | grep -oE '[0-9]+ file' | grep -oE '[0-9]+' || echo "0")
+  ```
+
+  Store `drift_detected=true`, `commit_count`, `changed_count` for dashboard rendering.
+
+  If `baseline_commit` is empty or matches `current_commit`: set `drift_detected=false`.
+
+  Restore identically regardless of time elapsed — no warnings about session age.
+
+  ---
+
+  ### Step 6: Compute Workflow Position and Next-Step Guidance
+
+  Compute `suggested_next` dynamically from COLONY_STATE.json data. Do not use the static value from session.json.
+
+  Use this decision tree:
+
+  ```
+  Case 1 — No plan created yet:
+    Check: plan.phases is empty AND plan.generated_at is null
+    recommended = "/ant:plan"
+    reason = "No plan created yet"
+    alternatives = ["/ant:colonize — analyze codebase first"]
+
+  Case 2 — Plan ready, first phase not started:
+    Check: plan.phases is not empty AND state == "READY" AND current_phase == 0
+    recommended = "/ant:build 1"
+    reason = "Plan ready, first phase not started"
+    alternatives = ["/ant:plan — review or regenerate plan"]
+
+  Case 3 — Build in progress:
+    Check: state == "EXECUTING"
+    recommended = "/ant:continue"
+    reason = "Build in progress"
+    alternatives = ["/ant:build {current_phase} — rebuild current phase", "/ant:flags — check for blockers"]
+
+  Case 4 — Phase complete, next phase available:
+    Check: state == "READY" AND current_phase > 0 AND current_phase < plan.phases.length
+    next = current_phase + 1
+    recommended = "/ant:build {next}"
+    reason = "Phase {current_phase} complete, ready for next"
+    alternatives = ["/ant:plan — regenerate plan", "/ant:phase {next} — preview next phase"]
+
+  Case 5 — All phases complete:
+    Check: state == "READY" AND current_phase > 0 AND current_phase >= plan.phases.length
+    recommended = "/ant:seal"
+    reason = "All phases complete"
+    alternatives = ["/ant:status — view final state"]
+
+  Case 6 — Colony paused:
+    Check: state == "PAUSED"
+    recommended = "/ant:resume-colony"
+    reason = "Colony is paused"
+    alternatives = ["/ant:status — check state first"]
+
+  Default:
+    recommended = "/ant:status"
+    reason = "Check colony status"
+    alternatives = []
+  ```
+
+  ---
+
+  ### Step 7: Workflow-Step Blocking (Early-Return Guards)
+
+  Run these guards BEFORE rendering the dashboard. If a blocking condition is detected, output the block message and STOP. Do not render the dashboard. Do not offer alternative commands.
+
+  **BLOCK CONDITION 1: No plan exists**
+
+  Check: plan.phases is empty AND plan.generated_at is null
+
+  Output and STOP:
+
+  ```
+  BLOCKED: No plan exists yet.
+  Required: Run /ant:plan to create a build plan.
+  Goal: {goal}
+  ```
+
+  Stop here — do not continue to Step 8 or render the dashboard.
+
+  ---
+
+  **BLOCK CONDITION 2: Plan attempted but failed**
+
+  Check: plan.phases is empty AND plan.generated_at is not null
+
+  Output and STOP:
+
+  ```
+  BLOCKED: Plan was attempted but has no phases.
+  Required: Run /ant:plan to regenerate the plan.
+  Goal: {goal}
+  ```
+
+  Stop here — do not continue to Step 8 or render the dashboard.
+
+  ---
+
+  **BLOCK CONDITION 3: Build interrupted**
+
+  Check: state == "EXECUTING" AND the last 3 events show no recent build activity
+
+  Output and STOP:
+
+  ```
+  BLOCKED: Build may have been interrupted.
+  Required: Run /ant:continue to check and advance.
+  Goal: {goal}
+  ```
+
+  Stop here — do not continue to Step 8 or render the dashboard.
+
+  ---
+
+  ### Step 8: Render Dashboard
+
+  Lead with the next-step recommendation. Context follows underneath ("straight to action" ordering).
+
+  ```
+  {{#claude}}━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━{{/claude}}
+  {{#opencode}}═══════════════════════════════════════════════════{{/opencode}}
+  RESUME SESSION
+
+  Next: {recommended}
+        {reason}
+  {if alternatives exist:}
+  Also: {alternatives, comma-separated}
+  {end}
+
+  {if drift_detected:}
+  Note: Codebase changed since last session ({commit_count} commit(s), {changed_count} file(s) modified)
+  {end}
+
+  Goal: {goal}
+  State: {state}
+  Phase: {current_phase}/{total_phases}
+
+  Phase Progress:
+  {for each phase in plan.phases:}
+    [{status_icon}] Phase {id}: {name}
+  {end}
+  ```
+
+  Status icons:
+  - completed: `v` (checkmark)
+  - in_progress: `~` (tilde)
+  - pending: ` ` (space)
+
+  ```
+  {if memory.decisions is not empty:}
+  Recent Decisions:
+  {for each of the last 5 decisions:}
+    - {decision text}
+  {end}
+  {end}
+
+  {if signals array from Step 3 is not empty:}
+  Active Signals:
+  {{#claude}}
+  {for each focus signal:}
+    FOCUS: {focus text}
+  {end}
+  {for each constraint signal:}
+    REDIRECT: {constraint text}
+  {end}
+  {{/claude}}
+  {{#opencode}}
+  {for each signal in signals:}
+    {signal.type}: "{signal.content}" [{signal.effective_strength * 100 | floor}%]
+  {end}
+  {{/opencode}}
+  {end}
+  ```
+
+  ---
+
+  ### Step 8.5: Display Memory Health (Secondary)
+
+  {{TOOL_PREFIX "Loading memory health..."}}:
+  ```bash
+  bash .aether/aether-utils.sh resume-dashboard
+  ```
+
+  Extract memory_health from the JSON result:
+  - wisdom_count
+  - pending_promotions
+  - recent_failures
+
+  Display after the main dashboard:
+  ```
+  📊 Memory Health
+     Wisdom: {wisdom_count} entries | Pending: {pending_promotions} promotions | Failures: {recent_failures} recent
+
+     Run /ant:memory-details for full breakdown
+  ```
+
+  If all counts are 0, show:
+  ```
+  📊 Memory Health
+     No accumulated wisdom yet. Complete phases to build colony memory.
+  ```
+
+  Last Command: {last_command}
+  Session: {session_id}
+  ```
+
+  ---
+
+  ### Step 9: Mark Session Resumed
+
+  {{TOOL_PREFIX "Marking session as resumed..."}}:
+  ```bash
+  bash .aether/aether-utils.sh session-mark-resumed
+  ```
+
+  ### Step 10: Next Up
+
+  Generate the state-based Next Up block by {{TOOL_PREFIX "Generating Next Up suggestions..."}}:
+  ```bash
+  state=$(jq -r '.state // "IDLE"' .aether/data/COLONY_STATE.json)
+  current_phase=$(jq -r '.current_phase // 0' .aether/data/COLONY_STATE.json)
+  total_phases=$(jq -r '.plan.phases | length' .aether/data/COLONY_STATE.json)
+  bash .aether/aether-utils.sh print-next-up "$state" "$current_phase" "$total_phases"
+  ```
+
+  ---
+
+  ## Error Handling Reference
+
+  | Condition | Response |
+  |-----------|----------|
+  | session.json missing (exists=false) | "No previous session found" — offer /ant:init and /ant:status |
+  | COLONY_STATE.json missing or corrupted | Pause, ask user: start fresh or recover |
+  {{#claude}}
+  | constraints.json missing | Skip silently (no pheromones) |
+  {{/claude}}
+  {{#opencode}}
+  | pheromone-read fails | Skip silently (no pheromones) |
+  {{/opencode}}
+  | CONTEXT.md missing | Fall back to COLONY_STATE.json narrative |
+  | No plan phases, no generated_at | BLOCK — redirect to /ant:plan |
+  | Plan attempted but no phases | BLOCK — redirect to /ant:plan |
+  | State EXECUTING, events show no activity | BLOCK — redirect to /ant:continue |
+  | baseline_commit matches current HEAD | No drift warning shown |
+  | baseline_commit differs from current HEAD | Show informational drift note |
+
+  ---
+
+  ## Key Constraints
+
+  {{#claude}}
+  - Use Read tool for COLONY_STATE.json and constraints.json (not bash cat/jq).
+  {{/claude}}
+  {{#opencode}}
+  - Use Read tool for COLONY_STATE.json (not bash cat/jq). Use Bash tool for pheromone-read (applies decay calculation).
+  {{/opencode}}
+  - Use Bash tool only for aether-utils.sh commands and git commands
+  - Handle ALL missing/corrupted file cases gracefully
+  - Time-agnostic: restore identically regardless of how long ago the session was
+  - Decisions shown as flat list — no user vs Claude distinction
+  - Blocking guards run BEFORE dashboard rendering (early-return pattern)
+  - Drift detection is informational only — not alarming, not a blocker

package/.aether/commands/run.yaml

@@ -0,0 +1,193 @@
+name: "ant:run"
+description: "Autopilot — builds, verifies, learns, and advances through phases automatically with smart pausing"
+description_opencode: "🤖🐜🔄🐜🤖 Autopilot — builds, verifies, learns, and advances through phases automatically with smart pausing"
+body: |
+  You are the **Queen**. Execute `/ant:run` — the adaptive autopilot loop.
+
+  The arguments are: `{{ARGUMENTS}}`
+
+  ## Purpose
+
+  This command automates the build-continue-advance cycle across multiple phases.
+  It reads and executes the same playbooks used by `/ant:build` and `/ant:continue`,
+  chaining them in a loop with intelligent pause conditions.
+
+  ## Rules
+
+  1. Do **not** invoke nested slash commands (`/ant:build`, `/ant:continue`, etc.).
+  2. Use the Read tool to load each playbook file, then execute it inline.
+  3. Preserve variables/results from prior stages and pass them forward.
+  4. Stop immediately on any pause condition (defined below).
+  5. Keep existing behavior and output format from the playbooks.
+  6. Log `autopilot_advance` events after each successful phase transition.
+
+  ## Arguments
+
+  Parse `{{ARGUMENTS}}` for:
+  - `--max-phases N` — Max phases to process (default: all remaining)
+  - `--replan-interval N` — Pause for replan suggestion every N phases (default: 2)
+  - `--continue` — Resume after a replan pause without replanning
+  - `--dry-run` — Preview plan without executing
+  - `--verbose` / `-v`, `--no-visual`, `--no-suggest` — Pass through to playbooks
+
+  ```
+  /ant:run                       Run all remaining phases
+  /ant:run --max-phases 2        Run at most 2 phases then stop
+  /ant:run --replan-interval 3   Suggest replan every 3 phases instead of 2
+  /ant:run --continue            Resume after replan pause without replanning
+  /ant:run --dry-run             Preview the autopilot plan
+  /ant:run --max-phases 3 -v     Run 3 phases with verbose output
+  ```
+
+  ## Dry Run Mode
+
+  If `--dry-run`: read COLONY_STATE.json, list remaining incomplete phases
+  (applying `--max-phases` cap), display the plan, then stop without executing.
+
+  ```
+  ━━━ A U T O P I L O T   P R E V I E W ━━━
+  Goal: {goal} | Current: Phase {N} | Remaining: {count} | Max: {max or "all"}
+
+    Phase {id}: {name} ({task_count} tasks) -> build -> continue -> advance
+    ...
+
+  Pause triggers: test failures, critical Chaos findings, new blockers,
+  security gate failures, quality gate failures, runtime verification needed,
+  replan suggestion (every {replan_interval} phases)
+  ```
+
+  ## Autopilot Loop
+
+  ### Step 0: Initialize
+
+  1. Read `.aether/data/COLONY_STATE.json`; validate goal + plan.phases exist
+  2. Determine remaining incomplete phases; apply `--max-phases` cap
+  3. Set `phases_completed = 0`, `autopilot_start = $(date +%s)`
+  4. Record pre-build blocker count: `bash .aether/aether-utils.sh flag-check-blockers {phase}`
+  5. Display: `AUTOPILOT ENGAGED | Goal: {goal} | Phase {N} | Max: {max or "all"}`
+
+  ### Step 1: Build Phase
+
+  Execute build playbooks in order, carrying cross-stage state
+  (`phase_id`, `visual_mode`, `verbose_mode`, `suggest_enabled`,
+  `prompt_section`, `wave_results`,
+  `verification_status`, `synthesis_status`, `next_action`):
+
+  1. `.aether/docs/command-playbooks/build-prep.md`
+  2. `.aether/docs/command-playbooks/build-context.md`
+  3. `.aether/docs/command-playbooks/build-wave.md`
+  4. `.aether/docs/command-playbooks/build-verify.md`
+  5. `.aether/docs/command-playbooks/build-complete.md`
+
+  Capture the synthesis result for pause evaluation.
+
+  ### Step 2: Build Pause Check
+
+  **PAUSE if ANY of these are true** (display reason, log event, STOP):
+
+  | # | Condition | Source |
+  |---|-----------|--------|
+  | 1 | Watcher `verification_passed == false` | build-verify synthesis |
+  | 2 | Any Chaos finding severity `critical` or `high` | build-verify synthesis |
+  | 3 | New blocker flags created (count increased since Step 0.4) | flag-check-blockers |
+
+  On pause, display the AUTOPILOT PAUSED banner with reason, affected phase,
+  specific issues, and instruction: "Fix issues, then run /ant:run to resume."
+  Log: `"<timestamp>|autopilot_paused|run|Paused at Phase {id}: {reason}"`
+
+  **If no pause:** proceed to Step 3.
+
+  ### Step 3: Continue (Verify + Advance)
+
+  Execute continue playbooks in order, carrying cross-stage state
+  (`visual_mode`, `state`, `current_phase`, `verification_results`,
+  `gate_results`, `advancement_result`, `next_phase_id`, `completion_state`):
+
+  1. `.aether/docs/command-playbooks/continue-verify.md`
+  2. `.aether/docs/command-playbooks/continue-gates.md`
+  3. `.aether/docs/command-playbooks/continue-advance.md`
+  4. `.aether/docs/command-playbooks/continue-finalize.md`
+
+  **Autopilot override for runtime verification (Step 1.11 in continue-gates):**
+  Skip the AskUserQuestion prompt. Instead, auto-PAUSE with reason
+  "Runtime verification required" so the user can test manually before resuming.
+
+  ### Step 4: Continue Pause Check
+
+  **PAUSE if ANY of these are true:**
+
+  | # | Condition | Source |
+  |---|-----------|--------|
+  | 4 | Verification loop reported NOT READY | continue-verify |
+  | 5 | Gatekeeper found critical CVEs | continue-gates |
+  | 6 | Auditor critical findings or score < 60 | continue-gates |
+  | 7 | Unresolved blocker flags remain | continue-gates |
+  | 8 | Runtime verification needed | continue-gates Step 1.11 |
+  | 9 | All phases complete (no next phase) | continue-advance |
+  | 10 | Replan trigger fires (unless `--continue`) | autopilot-check-replan (Step 5.5) |
+
+  For condition 9: jump to Step 6 (celebration). For condition 10: PAUSE with
+  replan suggestion (see Step 5.5). For all others: PAUSE with reason, log event, STOP.
+
+  **If no pause:** proceed to Step 5.
+
+  ### Step 5: Auto-Advance and Loop
+
+  1. Increment `phases_completed`
+  2. Update autopilot state: `bash .aether/aether-utils.sh autopilot-update --action advance --phase {next} --result success`
+  3. Log: `"<timestamp>|autopilot_advance|run|Phase {prev} -> {next} ({phases_completed}/{max})"`
+  4. Display: `--- Autopilot: Phase {prev} done -> Phase {next} ({N}/{max}) ---`
+  5. **Replan check** (see Step 5.5)
+  6. If `phases_completed >= max_phases` or no incomplete phases: go to Step 6
+  7. Otherwise: update `current_phase`, return to Step 1
+
+  ### Step 5.5: Replan Trigger Check
+
+  After each successful phase advance, check if a replan pause should fire:
+
+  ```bash
+  bash .aether/aether-utils.sh autopilot-check-replan --interval {replan_interval}
+  ```
+
+  If `--continue` flag was passed: skip this check entirely (user dismissed replan).
+
+  If `result.should_replan == true`: **PAUSE** with replan suggestion banner:
+
+  ```
+  ━━━ R E P L A N   S U G G E S T E D ━━━
+  Phases auto-completed: {N} | Learnings accumulated: {learnings_since_last}
+
+  The colony has completed {N} phases since the last plan review.
+  New learnings may have changed the optimal path forward.
+
+  Options:
+    /ant:plan              Regenerate phases with current learnings
+    /ant:run --continue    Dismiss and continue without replanning
+  ```
+
+  Log: `"<timestamp>|autopilot_replan_pause|run|Replan suggested after {N} phases ({learnings} learnings)"`
+
+  If `result.should_replan == false`: proceed normally (no pause).
+
+  ### Step 6: Final Summary
+
+  ```
+  ━━━ A U T O P I L O T   C O M P L E T E ━━━
+  Phases completed: {N} | Elapsed: {Xm Ys} | Now at: Phase {current}
+
+  {all complete}  -> Colony goal achieved! Run /ant:seal
+  {max reached}   -> Run /ant:run to continue
+  {replan}        -> Run /ant:plan to replan, or /ant:run --continue to dismiss
+  {paused}        -> Fix {reason}, then /ant:run to resume
+  ```
+
+  Update session:
+  `bash .aether/aether-utils.sh session-update "/ant:run" "/ant:run" "Autopilot: {N} phases, now Phase {current}"`
+
+  ## Execution Contract
+
+  For each playbook stage: Read file, execute inline, track `{stage_name, status, key_outputs}`.
+  If `status == failed`: evaluate pause conditions. Pause = graceful stop with saved state.
+  Hard failure (e.g., state corruption) = halt immediately, no recovery attempt.
+
+  On every pause: save COLONY_STATE.json, log event, update session, display resume instructions.