| name | virtual-scott-tick |
| description | Autonomous tick — invoked by the OpenClaw cron job every 5 minutes. Runs the deterministic Python orchestrator that discovers replyable comments and dispatches the reply skill for each. Not meant to be invoked manually outside of `openclaw cron run virtual-scott-tick` or `make poll-now`. |
| user-invocable | false |
Autonomous tick
Cron-fired entry point for the autonomous Virtual Scott loop. This
skill's whole job is to delegate to the deterministic Python
orchestrator at
~/.openclaw/workspace-virtual-scott/scripts/vs-tick.py.
The orchestrator handles: kill-switch check, WP REST discovery,
filtering, rate-limit cap, per-candidate dispatch to a child agent
running virtual-scott-reply, and success/failure logging into
runs/successes.jsonl / runs/failures.jsonl.
Why the logic lives in Python, not in this skill body: LLM-driven
orchestration of an N-iteration loop with branching is unreliable.
Deterministic shell/Python is the right tool for the loop layer; the
LLM is the right tool for the per-comment reasoning layer. Each per-
comment reply IS a full agent invocation of virtual-scott-reply —
so the autonomous behavior is still openclaw-native, just with a thin
deterministic conductor.
What to do
Run exactly one tool call. That's the entire skill.
The script's stdout is human-readable; report it verbatim in your
reply to the user.
If the exit code is non-zero, also report stderr. Do not retry. Do
not try to do the orchestration in your own reasoning instead of
running the script.
Kill switch
If the file ~/.openclaw/workspace-virtual-scott/PAUSED exists, the
script prints a "PAUSED" line and exits 0 without doing anything.
Pause: touch ~/.openclaw/workspace-virtual-scott/PAUSED. Resume:
rm ~/.openclaw/workspace-virtual-scott/PAUSED. The Makefile
exposes make pause and make resume.
Configuration
Two env vars, both optional, both honored by the Python script:
VS_MAX_REPLIES_PER_CYCLE (default 1): cap on replies per tick.VS_AGENT_TIMEOUT_SEC (default 600): timeout per child agent
invocation.
These can be set in the cron job's env (via OpenClaw cron config)
or directly when running make poll-now.
