feat(skills): add walking-mode-monitor skill for dog-walk QA sessions#414
Open
Nathan Schram (nathanschram) wants to merge 2 commits into
Open
feat(skills): add walking-mode-monitor skill for dog-walk QA sessions#414Nathan Schram (nathanschram) wants to merge 2 commits into
Nathan Schram (nathanschram) wants to merge 2 commits into
Conversation
New Claude Code skill that encodes a monitoring-only QA workflow run during dog walks (1-2 hours, 15-min ticks) on lba-1 VPS. Each tick reads Untether staging logs + Claude Code session logs, categorises observations (bug / blocked-tool / inefficiency / ux-ui / upstream), dedupes against session state + existing GitHub issues, and files new issues with label `auto:walking-mode` and full context. The skill is strictly monitoring-and-filing-only. Three overlapping guardrails: 1. Claude Code tool allow-list (frontmatter: Read, Glob, Grep, Bash, WebFetch) — harness rejects Edit/Write/NotebookEdit before Claude can invoke them. 2. Bash sub-command allow/deny lists documented in the skill body — explicit deny on git-mutating commands, gh pr merge/release, systemctl restart/stop/start, launchctl, sudo, package installers. 3. Prompt-boilerplate INVARIANT reminder at the top of every tick. Outer shell scheduler `~/.local/bin/walking-mode-monitor.sh` (not in repo — lba-1 operator config) spawns `claude -p` cold each tick, `timeout 600` per tick, state in `~/.untether-walking-mode/<id>/`. Calibration: first 2-3 walks run soft categories (inefficiency, ux-ui, upstream) in flag-only mode via `$STATE_DIR/flag-only-all` marker; hard categories (bug, blocked-tool) file immediately. GitHub label `auto:walking-mode` created separately. Sibling to existing `auto:error-report` label used by untether-issue-watcher for deterministic error events; walking-mode is complementary and cross-dedupes against it. Plan: ~/.claude/plans/untether-you-are-running-snoopy-quail.md Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
|
Important Review skippedAuto reviews are disabled on base/target branches other than the default branch. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
…itoring reference Key change: walking-mode-monitor no longer needs an outer shell script. The skill now handles its own bootstrap (state dir + session.json on first invocation) and self-pacing (ScheduleWakeup between ticks; no ScheduleWakeup on the wrap-up tick). `claude /loop` is the scheduler. Shell script ~/.local/bin/walking-mode-monitor.sh is retired (renamed .deprecated on lba-1). SKILL.md changes: - Frontmatter allowed_tools now includes ScheduleWakeup. - Invocation is `claude /loop "start walking-mode for 90 min..."`. - Bootstrap step added: first invocation creates state dir + session.json. - Per-tick step 7 now self-schedules: ScheduleWakeup if >15 min left, else run wrap-up directly. - Wrap-up now archives state and exits without ScheduleWakeup (loop terminates naturally). - Dry-run / flag-only / stop / error-handling all use the new flow. New reference doc `docs/reference/automated-monitoring.md`: - Pattern-level quick-ref for all Claude Code observe-and-file sessions. - Walking-mode is the first example; structured to accept future sessions (daily security audit, nightly cost digest, release smoke). - Covers shared invariants, invocation, how to add new sessions, troubleshooting, interaction with untether-issue-watcher, cost notes. CLAUDE.md updates: - "Automated monitoring" row added to Reference docs table. - Walking-mode row added to Skills table. Plan: ~/.claude/plans/untether-you-are-running-snoopy-quail.md Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
New Claude Code skill
walking-mode-monitorthat encodes a monitoring-only QA workflow for Untether staging. Invoked every 15 min during dog walks via an outer shell script on lba-1. The skill reads logs, categorises observations, dedupes against session state + existing GH issues, and files new issues with labelauto:walking-mode.Three-layer "monitoring-only" guardrail
[Read, Glob, Grep, Bash, WebFetch].Edit/Write/NotebookEdit/Taskrejected at the harness level.git push|merge|commit,gh pr merge|release,systemctl restart|stop|start,launchctl,sudo, package installers.What ships in this PR
.claude/skills/walking-mode-monitor/SKILL.md— canonical skill doc with all three guardrails, tick workflow, dedup algorithm, GH issue template with secret-redaction rules, wrap-up + dry-run + flag-only modes, Bash allow/deny reference, self-check checklist.What ships outside this PR (lba-1 operator config)
~/.local/bin/walking-mode-monitor.sh— outer shell scheduler. Not in repo (intentional — it's lba-1-local operator config, likeuntether-issue-watcher).auto:walking-mode— already created viagh label create(one-off).How it's invoked
Each tick cold-starts
claude -p "Walking-mode tick N for session <id>...". Cold start avoids context bloat; dedup lives in~/.untether-walking-mode/<id>/session.json.timeout 600per tick caps runaway cost.Calibration plan
First 2–3 walks: soft categories (
inefficiency,ux-ui,upstream) run in flag-only mode (logged but not filed). Hard categories (bug,blocked-tool) file immediately. Promote after signal quality confirmed.Interaction with existing tooling
untether-issue-watcher(which filesauto:error-reportissues for deterministic error events). Walking-mode handles judgment-required observations. Skill dedupes across both labels.src/**ortests/**impact.Plan
~/.claude/plans/untether-you-are-running-snoopy-quail.mdhas the full context + design rationale.Test plan
auto:walking-modecreatedbash -n) and clamp-boundary-tested🤖 Generated with Claude Code