OpenClaw ReadyMode Bot — Runtime Testing Guide¶

Comprehensive runtime testing manual for the ReadyMode Support Discord bot. Covers Discord-side and server-side test paths, the full T1–T18 battery, troubleshooting, and known non-bugs.

1. Overview¶

ReadyMode Support is an LLM-driven Discord bot that automates operator-level tasks inside the ReadyMode CRM web app via a Chrome headless instance controlled over CDP.

Field	Value
Discord display name	`ReadyMode Support`
Discord user ID	`1491866571818139718`
Guild ID	`1476748033134956756`
Channel ID	`1491870539713609848`
ReadyMode account	`readymode` (server-side process user `agent`)
Pivot domain	`breaddialer.readymode.com` (Arpa Growth suspended since April 2026)
Admin login flag	`READYMODE_LOGIN_AS_ADMIN=1` → logs in as shared `manager` account
Chrome CDP port	`18800`
Host	`159.89.179.179` (SSH on port `22022`)

What the bot does¶

The bot reads a Discord message, decides via its LLM + the SKILL.md decision table, and either:

Replies conversationally (smoke, KB queries, ambiguous requests),
Invokes a bash script in /home/agent/.openclaw/workspace-readymode/skills/readymode-support/scripts/ that drives Chrome.

Scripts available¶

Script	Purpose	Destructive?
`create_user.sh`	Create operator user; assign states + campaigns	Yes (real)
`clear_licenses.sh`	Wipe license assignments for a user	Yes
`upload_leads.sh`	Upload a CSV of leads into a campaign	Yes
`reset_leads.sh`	NOT IMPLEMENTED — bot replies "no automatizado" + phone	n/a

Where things live on the server¶

/home/agent/.openclaw/
├── workspace-readymode/
│   └── skills/readymode-support/
│       ├── SKILL.md                  # decision table (LLM prompt)
│       ├── scripts/*.sh              # automation scripts
│       └── screenshots/*.png         # 73-byte pointer files (see §5)
├── media/browser/<uuid>.png          # real PNG payloads
├── state/readymode-halt.flag         # circuit-breaker flag
└── logs/readymode/*.jsonl            # session logs

2. Testing in Discord¶

2.1 Joining the bot's channel¶

Accept the guild invite for guild 1476748033134956756 (ask Miguel for the invite link if you don't have access yet).
Open channel #readymode-support — channel ID 1491870539713609848.
Confirm the bot's presence dot is green in the member sidebar.

2.2 Channel configuration¶

The bot is configured with requireMention: false, which means:

You do not need to @ReadyMode Support to address it.
Any message dropped in the channel is consumed.
Do not write off-topic chatter there — it will be parsed as intent.

2.3 Phrases the bot understands¶

Pulled from the SKILL.md decision table. Both ES and EN forms are accepted, but the bot tends to reply in Spanish (see §6).

Intent	Spanish phrasing	English phrasing
Smoke / capabilities	"hola", "qué puedes hacer", "ayuda"	"hi", "what can you do", "help"
Create user	"crea usuario `<name>` estados `<X,Y>` campañas `<C>`"	"create user `<name>` states ... campaigns ..."
Clear licenses	"limpia licencias de `<user>`", "clear licenses"	"clear licenses for `<user>`"
Upload leads	"sube los leads a `<campaign>`" (+ CSV attached)	"upload leads to `<campaign>`"
Reset leads	"reset leads de `<user>`"	"reset leads for `<user>`"
KB query (support)	"línea de soporte", "horario", "contacto"	"support line", "hours"

2.4 Smoke test: is the bot alive?¶

Send in the channel:

hola dime qué puedes hacer

Expected: a short Spanish reply listing create user / clear licenses / upload leads / reset leads / support contact. Typical latency: 2–6 s.

If no reply in 30 s, check the HALT flag (§5.1) and the gateway health (§3.3).

2.5 Monitoring bot logs from SSH¶

ssh -p 22022 -i ~/.ssh/id_rsa miguel@159.89.179.179

# Tail the current readymode session log
sudo -n -u agent bash -c 'ls -1t /home/agent/.openclaw/logs/readymode/*.jsonl | head -1 | xargs -I{} tail -F {}'

# Pretty-print one line
sudo -n -u agent bash -c 'tail -1 /home/agent/.openclaw/logs/readymode/*.jsonl' | jq .

JSONL fields of interest: ts, kind (user_msg | tool_call | script_stdout | assistant_msg), script, exitCode, stderr.

3. Testing on the server (CLI / TUI)¶

3.1 SSH in¶

ssh -p 22022 -i ~/.ssh/id_rsa miguel@159.89.179.179

SSH port is 22022 (NHE-29 migrated it off 22). If you get a timeout, run nc -zv 159.89.179.179 22022 before blaming firewall.

3.2 One-shot agent message (does NOT round-trip through Discord)¶

Useful for fast iteration without polluting the channel:

sudo -n -u agent bash -c "cd /home/agent && \
  /usr/bin/node /home/agent/.npm-global/lib/node_modules/openclaw/openclaw.mjs \
  agent --agent readymode --message \"hola dime qué puedes hacer\""

Replace the quoted message for each test. The reply prints to stdout. The session is ephemeral — no multi-turn state is preserved across invocations.

3.3 Multi-turn TUI¶

For tests that require confirmation prompts (T5, T6) or chained turns:

sudo -n -u agent bash -c "/usr/bin/node /home/agent/.npm-global/lib/node_modules/openclaw/openclaw.mjs \
  tui --session readymode-manual-$(date +%s)"

Each invocation gets a unique session key so logs are easy to find.

3.4 Gateway health¶

sudo -n -u agent bash -c "/usr/bin/node /home/agent/.npm-global/lib/node_modules/openclaw/openclaw.mjs health"

Expected output: Discord gateway connected, plugin readymode-support loaded, Chrome CDP reachable on localhost:18800.

4. Test battery T1–T18¶

Legend: - ✅ already validated today (2026-05-12) - ⚠️ known imperfect behavior (see §6) - ⏳ pending multi-turn / Discord round-trip - 🔴 destructive on the real CRM — coordinate with Miguel before running

T1 — Smoke ✅¶

Field	Value
Message	`hola dime qué puedes hacer`
Expected	Short Spanish capability list (create user / clear licenses / upload leads / reset leads / support contact)
Verify	Reply within ~6 s; no script invocation in JSONL

T2 — Create user, single state ✅¶

Field	Value
Message	`crea usuario testbot04 estados TX campañas Main`
Expected	"Usuario `testbot04` creado" + confirmation of states/campaigns
Verify	JSONL contains `tool_call` → `create_user.sh` with `exitCode:0`; Chrome screenshot in `/home/agent/.openclaw/media/browser/`

T3 — Create user, multi-state, auto-kick ✅¶

Field	Value
Message	`crea usuario testbot05 estados CA,TX campañas Main`
Expected	User created. Bot will auto-kick any human session on `manager` (logged warning, then proceeds)
Verify	JSONL line `kick_event:true`; final `exitCode:0`. Warn Miguel before running if he's actively logged in.

T4 — Create user without params ✅¶

Field	Value
Message	`crea usuario`
Expected	Clarification prompt asking for name + states + campaigns. No script call.
Verify	No `tool_call` entry in JSONL

T5 — Clear licenses → confirm yes ⏳ 🔴¶

Requires multi-turn (TUI or Discord). Two messages:

> limpia licencias de testbot04
< ¿Confirmas limpiar todas las licencias de testbot04? (sí/no)
> si
< Limpiando…  → Listo.

Verify: clear_licenses.sh invoked with exitCode:0. Screenshot of licenses page post-clear.

T6 — Clear licenses → confirm no ⏳¶

Same start, but reply no:

> limpia licencias de testbot04
< ¿Confirmas limpiar todas las licencias de testbot04? (sí/no)
> no
< Ok, cancelo la operación.

Verify: no tool_call for clear_licenses.sh. Final assistant message is a polite cancel.

T7 — Upload leads with CSV attachment ⏳ 🔴¶

Only testable via Discord (CSV upload). Use the prepared file at /home/apolo/leads_test.csv.

Attach the CSV to the next Discord message.
Send: sube los leads a Main
Expected: Bot acknowledges, runs upload_leads.sh, returns row count + campaign name.

Verify: JSONL upload_leads.sh exitCode:0; ReadyMode campaign view shows the new leads.

T8 — Upload leads, no attachment ✅¶

Field	Value
Message	`sube los leads a Main`
Expected	"Necesito un CSV adjunto" (or similar) — no script call
Verify	No `tool_call` in JSONL

T9 — KB query: hours ⚠️¶

Field	Value
Message	`¿cuál es el horario de soporte?`
Expected (KB verbatim)	`Lun–Vie 6am–6pm PT`
Actual today	LLM paraphrases; sometimes adds extra context not in KB
Severity	Cosmetic — see §6

T10 — KB query: contact email ⚠️¶

Field	Value
Message	`email de soporte`
Expected (KB verbatim)	`support@readymode.com`
Actual today	LLM paraphrases ("puedes escribir a…")

T11 — KB query: ticket portal ⚠️¶

Field	Value
Message	`portal de tickets`
Expected	KB URL verbatim
Actual today	LLM improvises

T12 — KB query: escalation ⚠️¶

Field	Value
Message	`¿cómo escalo un problema?`
Expected	KB escalation steps verbatim
Actual today	LLM rewrites in its own words

T13 — Support line ✅¶

Field	Value
Message	`línea de soporte`
Expected (verbatim)	`1 (800) 694-1049 ext. 4`
Verify	Exact string appears in reply

T14 — Reset leads (unimplemented path) ✅¶

Field	Value
Message	`reset leads de Juan`
Expected	"Esa operación no está automatizada todavía. Llama al `1 (800) 694-1049 ext. 4`"
Verify	No `reset_leads.sh` invocation (script does not exist)

T15 — Invalid USPS state code ⚠️ 🔴¶

Field	Value
Message	`crea usuario testbot06 estados ZZ campañas Main`
Expected	Bot should reject `ZZ`
Actual today	`create_user.sh` proceeds; user is created without that state license. Not a regression — script lacks state validation (NHE-44 G14).

T16 — English input ⚠️¶

Field	Value
Message	`clear licenses for testbot04 please`
Expected	Bot understands and asks for confirmation in English
Actual today	Bot understands intent, replies in Spanish (cosmetic — §6)

T17 — Two intents in one message ⏳¶

Field	Value
Message	`crea usuario testbot07 estados TX campañas Main y limpia licencias de testbot05`
Expected	Bot should sequence: ask for clarification or execute one then prompt for the other.
Verify	Inspect JSONL — should see at most one `tool_call` before a clarifying message

T18 — HALT flag active ⏳¶

Pre-condition: HALT flag present.

sudo -n -u agent touch /home/agent/.openclaw/state/readymode-halt.flag

Then send any actionable command (e.g., crea usuario testbot99 estados TX campañas Main).

Expected: Bot replies with a maintenance/halt message and does not invoke any script.

Cleanup:

sudo -n -u agent rm /home/agent/.openclaw/state/readymode-halt.flag

5. Troubleshooting¶

5.1 HALT flag¶

A file at /home/agent/.openclaw/state/readymode-halt.flag short-circuits all destructive script paths.

# Check
sudo -n -u agent test -f /home/agent/.openclaw/state/readymode-halt.flag && echo "HALTED" || echo "ok"

# Clear
sudo -n -u agent rm -f /home/agent/.openclaw/state/readymode-halt.flag

When the flag is set, the bot still replies conversationally but refuses to drive Chrome.

5.2 "already logged in"¶

Symptom: script aborts with an already logged in modal error from ReadyMode.

Cause: shared manager account collision — a human (likely Miguel) is logged in elsewhere.

Fix: patch C2 auto-kicks the human session and retries the login. If the patch is active you'll see kick_event:true in the JSONL and the script proceeds. If you still see the error, the patch may have regressed — alert Miguel before forcing again.

Long-term: NHE-56 G22 (dedicated bot account) closes this class permanently.

5.3 "tomando tiempo inusualmente largo"¶

The bot LLM sometimes posts an "this is taking unusually long" message before the underlying script has actually finished. The script almost always completes successfully a few seconds later.

Verify by tailing the JSONL: if exitCode:0 arrives after the impatience message, the operation succeeded. Don't re-run.

5.4 Screenshots are 73-byte pointers¶

Files in /home/agent/.openclaw/workspace-readymode/skills/readymode-support/screenshots/*.png are 73-byte pointer stubs, not real PNGs. The actual image lives at /home/agent/.openclaw/media/browser/<uuid>.png.

# Find real PNG referenced by a pointer
sudo -n -u agent cat /home/agent/.openclaw/workspace-readymode/skills/readymode-support/screenshots/<name>.png
# → outputs e.g. "media/browser/<uuid>.png"

# Copy real one back to local
scp -P 22022 -i ~/.ssh/id_rsa \
  miguel@159.89.179.179:/home/agent/.openclaw/media/browser/<uuid>.png \
  /tmp/

5.5 Chrome CDP unreachable¶

If health reports CDP not reachable on 18800:

# Status
sudo -n -u agent bash -c "curl -sS http://localhost:18800/json/version" | jq .

# Restart Chrome — only with explicit Miguel approval (production bot is live)

Do not restart Chrome or any readymode service without an explicit go-ahead. The bot is live in production.

6. Known issues — NOT new bugs¶

Do not file these as regressions. They are tracked elsewhere or are accepted current behavior.

#	Issue	Status / Tracker
1	LLM paraphrases KB answers when multiple matches exist (T9–T12)	Prompt-tuning backlog
2	Bot replies in Spanish even when prompted in English (T16)	Style decision, deferred
3	`create_user.sh` does not validate USPS state codes; accepts `ZZ` (T15)	NHE-44 G14a-e
4	Playlist drag-and-drop for states/campaigns not implemented; assignments may end up pending	NHE-44 G14a-e
5	Screenshots in `skills/.../screenshots/` are 73-byte pointers, not PNGs	By design (§5.4)
6	"Tomando tiempo inusualmente largo" precedes successful script completion	LLM patience tuning

Appendix A — Quick command cheatsheet¶

# SSH in
ssh -p 22022 -i ~/.ssh/id_rsa miguel@159.89.179.179

# Health
sudo -n -u agent bash -c "/usr/bin/node /home/agent/.npm-global/lib/node_modules/openclaw/openclaw.mjs health"

# One-shot agent message
sudo -n -u agent bash -c "/usr/bin/node /home/agent/.npm-global/lib/node_modules/openclaw/openclaw.mjs agent --agent readymode --message \"hola\""

# Multi-turn TUI
sudo -n -u agent bash -c "/usr/bin/node /home/agent/.npm-global/lib/node_modules/openclaw/openclaw.mjs tui --session readymode-manual-$(date +%s)"

# Tail live JSONL
sudo -n -u agent bash -c 'ls -1t /home/agent/.openclaw/logs/readymode/*.jsonl | head -1 | xargs -I{} tail -F {}'

# HALT controls
sudo -n -u agent touch  /home/agent/.openclaw/state/readymode-halt.flag    # halt
sudo -n -u agent rm -f  /home/agent/.openclaw/state/readymode-halt.flag    # resume

Appendix B — Test status snapshot (2026-05-12)¶

ID	Status	Note
T1	✅	smoke OK
T2	✅	testbot04 created
T3	✅	testbot05 created with auto-kick
T4	✅	clarification path works
T5	⏳	needs Discord/TUI
T6	⏳	needs Discord/TUI
T7	⏳	needs Discord (CSV attach)
T8	✅	rejects missing attachment
T9	⚠️	KB paraphrased
T10	⚠️	KB paraphrased
T11	⚠️	KB paraphrased
T12	⚠️	KB paraphrased
T13	✅	verbatim "1 (800) 694-1049 ext. 4"
T14	✅	"no automatizado" + phone
T15	⚠️	invalid state accepted (NHE-44)
T16	⚠️	replies in Spanish
T17	⏳	two-intent path untested
T18	⏳	HALT flag path untested