sitrep

Generate a situational report from live system state. READ-ONLY — no writes, no fixes.

Trigger

User types /sitrep or asks for status, "what's going on?", "where are we?", or "catch me up".

Usage

/sitrep              # Full report (default)
/sitrep --brief      # Alerts only, one-line summary per section

Note: --since is not yet implemented. Use git log or bus tail manually for deltas.

Modes

Full mode (default)

Run all data collection sections and emit the complete formatted report.

Brief mode (--brief)

Skip detailed per-section output. Emit only:

• Machine identity line
• Branch + dirty count
• Bus entry count + blocker count
• Service status table
• Disk line with alert badge
• Alerts section (this is the primary output in brief mode)

When --brief is passed, the agent should print the alerts section first, then the one-line summaries, then exit. No process lists, no commit details, no bus message content.

Protocol

1. Gather state from multiple sources (parallel where possible).
2. Never fabricate output — run actual commands.
3. Never write to disk, post to bus, or kill processes.
4. If a command fails, report the failure inline.
5. Flag disk > 85% as WARN, > 92% as CRIT.
6. Ollama on a GPU host (Workstation) is expected UP; on non-GPU hosts flag as WARN if down.

Platform Detection

Detect platform before running commands. The fleet has three OS targets:

Machine	OS	Shell	Canonical $HOME
Workstation	Windows 11	git-bash + PowerShell	$HOME
Huxley	macOS 13	zsh	/Users/owner
remote-node (dormant since 2026-07-01)	macOS 15	zsh	$REMOTE_HOST

Use this guard pattern at the top of every /sitrep invocation:

import platform, os, sys

PLATFORM = platform.system().lower()  # "windows", "darwin", "linux"
IS_MSYS = os.environ.get("MSYSTEM") is not None or sys.platform == "msys"
HOME = os.path.expanduser("~")

All shell snippets below are written with platform branches.

---

Data Collection

Run every section. If a probe fails, report ERR: <message>.

1. Project Root Discovery

import subprocess, os, pathlib

def find_project_root():
    # 1. Cwd if inside a git repo
    try:
        out = subprocess.check_output(
            ["git", "rev-parse", "--show-toplevel"],
            stderr=subprocess.DEVNULL,
            text=True,
        ).strip()
        if out:
            return pathlib.Path(out)
    except Exception:
        pass
    # 2. CORPUS_ROOT env var
    if os.environ.get("CORPUS_ROOT"):
        return pathlib.Path(os.environ["CORPUS_ROOT"])
    # 3. PROJECTS_DIR env var + hummbl-governance
    base = os.environ.get("PROJECTS_DIR") or os.path.expanduser("~/PROJECTS")
    for candidate in ["hummbl-governance", "corpus"]:
        p = pathlib.Path(base) / candidate
        if (p / ".git").exists() or (p / "manifest.json").exists():
            return p
    return None

PROJECT_ROOT = find_project_root()

2. Git State

Windows (PowerShell via agent-powershell.ps1 or direct pwsh):

$branch = git -C "$PROJECT_ROOT" rev-parse --abbrev-ref HEAD 2>$null
if (-not $branch) { $branch = "unknown" }
git -C "$PROJECT_ROOT" log --oneline -5
git -C "$PROJECT_ROOT" status --short | Measure-Object -Line

Unix (bash/zsh):

branch=$(git -C "$PROJECT_ROOT" rev-parse --abbrev-ref HEAD 2>/dev/null || echo "unknown")
git -C "$PROJECT_ROOT" log --oneline -5
git -C "$PROJECT_ROOT" status --short | wc -l | tr -d ' '

3. Bus Activity

Scan multiple candidate paths. The canonical bus was retired 2026-04-21; live paths vary by machine.

Candidates (checked in order):

1. $CORPUS_ROOT/_state/coordination/messages.tsv
2. $PROJECT_ROOT/hummbl_governance/_state/coordination/messages.tsv
3. $HOME/.agents/bus/messages.tsv
4. $HOME/_state/coordination/messages.tsv
5. remote-node bridge: python $HOME/bin/bus-global.py list --limit 5 (if SSH/remote-node reachable — remote-node dormant since 2026-07-01, expected UNREACHABLE)

Windows:

$busPaths = @(
    "$env:CORPUS_ROOT\_state\coordination\messages.tsv",
    "$env:PROJECT_ROOT\hummbl_governance\_state\coordination\messages.tsv",
    "$HOME\.agents\bus\messages.tsv",
    "$HOME\_state\coordination\messages.tsv"
)
$foundBus = $null
foreach ($p in $busPaths) {
    if (Test-Path $p) { $foundBus = $p; break }
}
if ($foundBus) {
    (Get-Content $foundBus).Count
    Get-Content $foundBus | Select-Object -Last 5
} else {
    echo "Bus: no local TSV found"
}

Unix:

bus_paths=(
    "${CORPUS_ROOT:-}/_state/coordination/messages.tsv"
    "${PROJECT_ROOT:-}/hummbl_governance/_state/coordination/messages.tsv"
    "$HOME/.agents/bus/messages.tsv"
    "$HOME/_state/coordination/messages.tsv"
)
found_bus=""
for p in "${bus_paths[@]}"; do
    if [ -f "$p" ]; then found_bus="$p"; break; fi
done
if [ -n "$found_bus" ]; then
    wc -l < "$found_bus"
    tail -5 "$found_bus"
else
    echo "Bus: no local TSV found"
fi

Blockers: Search last 10 lines for BLOCKED type.

4. Services

Discovery approach: Check configurable ports via netstat / Get-NetTCPConnection / lsof. Do NOT assume hardcoded ports are correct on every machine.

Port map (read from env first, fallback to known defaults):

Service	Env Var	Default	Machine Context
Ollama	OLLAMA_HOST	localhost:11434	Workstation (GPU host) = expected; others = optional
Gitea	GITEA_URL	localhost:3030	Workstation only
remote-node bus bridge	BUS_BRIDGE	maks-mac-mini:18790	Mesh-wide (remote-node dormant since 2026-07-01 — use Workstation as bus authority)
Tailscale	—	tailscaled daemon	Mesh-wide

Windows probe:

function Test-Port($hostName, $port) {
    try {
        $conn = Test-NetConnection -ComputerName $hostName -Port $port -WarningAction SilentlyContinue
        if ($conn.TcpTestSucceeded) { "UP" } else { "DOWN" }
    } catch { "ERR" }
}
Test-Port "localhost" 11434
Test-Port "localhost" 3030
Test-Port "maks-mac-mini" 18790  # remote-node dormant since 2026-07-01 — expected DOWN

Unix probe:

test_port() {
    timeout 2 bash -c "</dev/tcp/$1/$2" 2>/dev/null && echo "UP" || echo "DOWN"
}
test_port localhost 11434
test_port localhost 3030
test_port maks-mac-mini 18790  # remote-node dormant since 2026-07-01 — expected DOWN

Ollama policy:

• If machine hostname is workstation OR nvidia-smi is available → Ollama expected UP. Flag as WARN if DOWN.
• Otherwise Ollama is optional. Flag as INFO if DOWN.

5. Running Agents / Processes

Windows:

Get-Process | Where-Object {
    $_.ProcessName -match "claude|codex|devin|gemini|node|python|ollama"
} | Select-Object ProcessName, Id, @{N="CPU(s)";E={[math]::Round($_.CPU,1)}}, @{N="WS_MB";E={[math]::Round($_.WorkingSet/1MB,1)}}

Unix:

ps aux | grep -iE "claude|codex|devin|gemini|node|python|ollama" | grep -v grep | awk '{print $11, $2, $3, $6}'

6. Disk Pressure

Windows:

Get-CimInstance Win32_LogicalDisk | Where-Object { $_.DeviceID -eq 'C:' } |
    Select-Object DeviceID,
        @{N="SizeGB";E={[math]::Round($_.Size/1GB,1)}},
        @{N="FreeGB";E={[math]::Round($_.FreeSpace/1GB,1)}},
        @{N="Used%";E={[math]::Round(($_.Size-$_.FreeSpace)/$_.Size*100,1)}}

Unix:

df -h / | tail -1 | awk '{print $5, $4}'

Alert thresholds:

• Used% < 70% → OK
• Used% 70–84% → OK (note)
• Used% 85–91% → WARN
• Used% >= 92% → CRIT

7. Skill / Fleet Surface (Optional but useful)

Count installed skills:

Windows:

$skillRoots = @("$HOME/.devin/skills", "$HOME/.claude/skills", "$HOME/.agents/skills")
$count = 0
foreach ($r in $skillRoots) {
    if (Test-Path $r) {
        # -Depth 1 prevents recursion into nested skill subdirectories
        $count += (Get-ChildItem -Directory $r -Depth 0).Count
    }
}
$count

Unix:

find "$HOME/.devin/skills" "$HOME/.claude/skills" "$HOME/.agents/skills" -maxdepth 1 -type d 2>/dev/null | wc -l

---

Output Format

SITREP | <YYYY-MM-DD HH:MM> <TZ>
Machine: <hostname> | <OS> | <platform>
==============================

## Git
Branch: <branch>
Project root: <path or "unknown">
Last 5 commits:
  <hash> <msg>
Dirty: <N files> [list if < 10]

## Coordination Bus
Local TSV: <path or "not found">
Entries: <N> total
Recent (last 5):
  <timestamp> <from> <to> <type> <msg_truncated>
Blockers: <N recent BLOCKED or "none">

## Services
| Service          | Target           | Status | Notes                |
|------------------|------------------|--------|----------------------|
| Ollama           | <host>:<port>    | UP/DOWN| expected/optional    |
| Gitea            | <host>:<port>    | UP/DOWN| Workstation-only           |
| Bus bridge       | <host>:<port>    | UP/DOWN| remote-node dormant since 2026-07-01 — Workstation is bus authority |
| Tailscale daemon | —                | UP/DOWN| mesh connectivity    |

Agents: <list of running agent processes or "none detected">

## Disk
<drive>: <Used%>% (<FreeGB> GB free of <SizeGB> GB) [OK/WARN/CRIT]

## Fleet Surface
Skills installed: <N>

## Alerts
- <one per line, or "All clear.">

Constraints

• READ-ONLY. Never fix, never write, never post.
• If Ollama is UP on agent-node → normal. If Ollama is DOWN on agent-node → WARN.
• If bus TSV is missing → report "not configured" rather than error.
• If git fails inside project root → report "not a git repo" and continue.
• Do not fabricate timestamps, commit hashes, or process counts.
• Keep output scannable in < 10 seconds.

Changelog

• 2026-06-15 v1.1 — Fixed Windows skill count recursion (-Depth 0). Added --brief mode protocol. Removed unimplemented --since from usage.
• 2026-06-15 v1.0 — Rewritten for cross-platform fleet (Windows/macOS/Linux). Replaced Unix-only commands with platform branches. Added bus path fallback chain. Added Ollama context-aware policy. Added project root auto-discovery.