Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add peteromallet/Arnold --skill pipeline-babysittinggit clone --depth 1 https://github.com/peteromallet/ArnoldWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/peteromallet/arnold/pipeline-babysitting)<a href="https://agentmods.dev/skills/peteromallet/arnold/pipeline-babysitting"><img src="https://agentmods.dev/badge/skills/peteromallet/arnold/pipeline-babysitting/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/peteromallet/arnold/pipeline-babysitting"><img src="https://agentmods.dev/badge/skills/peteromallet/arnold/pipeline-babysitting.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00232 | $0.03746 |
| Opus 5.5 | $0.00093 | $0.01498 |
| Sonnet 5.5 | $0.00046 | $0.00749 |
| Haiku 4.5 | $0.00023 | $0.00375 |
Grade A, and why
pipeline-babysitting scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 324 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Babysitting Principles
How we run the fix-the-fixer / babysitting loop, and what the harness should
aspire to. Canonical source: docs/pipeline-babysitting.md in the Arnold
repo; this skill is the agent-facing form (same content, plus the tools).
1. The hourly check-in
The babysitting loop runs as a once-per-hour check-in. Each check-in is
one pass over the live system: check → fix → push → re-arm. It never just
reports — every check-in ends with the run measurably closer to done or a
concrete blocker fixed at the root. (Acute stoppages between check-ins are
caught by the watchdog's status-trigger path — the arnold-babysitter
wrapper launches a single Flash agent whose goal is rendered by
skills/babysitter/scripts/render_babysitter_goal.py.)
1.1 Check for failures and repeated misses
- Pull the live status with the one-command tools (
fixer-status,epic-board) and read the living evidence doc (§3.6). - For every stopped/failed/stalled chain, ask: is this a repeated failure class the fixer isn't fixing, or isn't getting around to? Same failure fingerprint roughly three times in a row, or a chain blocked ~1 hour without autorecovery, is the trigger.
- This question targets the fixer's failure to fix, not the chain's failure to run.
1.2 The fixing loop — bias toward fixing the fixer
When something is stuck or a failure class repeats, run the loop entirely inside sub-agents:
- Understand — deploy a bounded, read-only swarm of DeepSeek Flash
sub-agents over the failure evidence (what happened, the code path, how
state is passed around) —
skills/subagent-launcher/fan.py. - Recommend — hand the packed context (evidence pack, swarm index, every investigator report) to a premium model — Grok — for the once-and-for-all recommendation.
- Implement — with DeepSeek Flash, fix the fixer itself (its prompt, its dispatch, its config) in the approved editable runtime — the chain-level fix is what the fixer does once it runs again. Verify against the focused regression.
- Relaunch the fixer — the deliverable of the loop is the fixer running
again with the fix in place: restart the fixer (and, when the evidence
requires, the chain via megaplan resume / chain start); never
--fresh. - Verify at the next check-in — the next hourly check-in checks whether
and how the fixer actually fixed the chain:
chain-*.jsonlast_stateleaves blocked and the samefailure_fingerprintdoes not recur. A PID, commit, self-report, or heartbeat is NOT proof.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago First seen · 324 lines · 232 tokens per session scan A 4960d2a91765
pipeline-babysitting is a skill published in the GitHub repository peteromallet/Arnold (98 stars, last pushed yesterday), licensed Apache-2.0. It adds 232 tokens to every session and 3,746 once invoked, about $0.0009 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-04.
Other skills, from other repositories
chinese-git-workflow
A reference for configuring Git with Chinese code-hosting services such as Gitee, Coding.net, GitLab China, and CNB, including SSH, HTTPS, credentials, CI, and repository mirroring.
atmos-profiles
Atmos profiles: profile directories, --profile and ATMOSPROFILE activation, profile merge behavior, environment switching, and routing profile-specific auth/toolchain/config overrides.
webhook-management
Configure and validate CCAM webhook targets across supported chat, incident, automation, and generic providers. Use when listing provider requirements, creating or updating a target, scoping it to alert rules, sending a test notification, reviewing delivery history, or deleting a target.
monorepo-management
Master monorepo management with Turborepo, Nx, and pnpm workspaces to build efficient, scalable multi-package repositories with optimized builds and dependency management. Use when setting up monorepos, optimizing builds, or managing shared dependencies.
managing-github-actions-secrets
Creates and updates GitHub Actions secrets for PostHog workflows. Use when adding a new CI secret, rotating an existing secret, wiring a workflow to an API token, package registry credential, deploy key, or any value referenced via ${{ secrets. }} in .github/workflows/.
deployment-engineer
Expert deployment engineer specializing in modern CI/CD pipelines, GitOps workflows, and advanced deployment automation. Masters GitHub Actions, ArgoCD/Flux, progressive delivery, container security, and platform engineering. Handles zero-downtime deployments, security scanning, and developer experience optimization.…