rtmalikian

LLM Tooling Agent Infra Quiet Failures — Data skill for Claude Code

Data community

When agent plumbing fails quietly: 11 open-source PRs to claude-mem, AnythingLLM, headroom and crawl4ai, 8 merged.

How to install LLM Tooling Agent Infra Quiet Failures

This entry records only its repository, not the path inside it, so there is no exact command to give. Open rtmalikian/llm-tooling-agent-infra-quiet-failures and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What LLM Tooling Agent Infra Quiet Failures does

When agent plumbing fails quietly: 11 open-source PRs to claude-mem, AnythingLLM, headroom and crawl4ai, 8 merged. Silent data loss, blind approvals, mislabelled errors, invented digits, and what clinical AI builders should check first.

Alternatives in Data

Sentry — Inspect Sentry issues, summarize production errors, and pull health data 14.6k ★
Opencode Mem — OpenCode plugin that gives coding agents persistent memory using local vector database 1.5k ★
Onboarding Validate — Run DBT tests and validate data quality; fix or flag errors 49 ★

README

Download the paper (PDF)

[![Download PDF](https://img.shields.io/badge/Download-PDF-red?style=for-the-badge&logo=adobeacrobatreader&logoColor=white)](https://github.com/rtmalikian/llm-tooling-agent-infra-quiet-failures/raw/main/week3-llm-tooling-agent-infra.pdf)

👉 [Click here to download the PDF](https://github.com/rtmalikian/llm-tooling-agent-infra-quiet-failures/raw/main/week3-llm-tooling-agent-infra.pdf)

One click, no sign-up, no paywall.

When Agent Plumbing Fails Quietly: A Retrospective Analysis of Open-Source Contributions to LLM Tooling and Agent Infrastructure

Raphael T. Malikian, MBBS BSc (Hons) — Independent Researcher, Palmdale, California, USA — ORCID 0009-0008-4326-4205

Submitted to Computer Methods and Programs in Biomedicine (Manuscript CMPB-D-26-08971). Preprint: Zenodo, https://doi.org/10.5281/zenodo.23741622.

**Running head:** WHEN AGENT PLUMBING FAILS QUIETLY

**Author:** Raphael T. Malikian, MBBS BSc (Hons)

**Affiliation:** Independent Researcher, Palmdale, California, USA

**Corresponding author:** Raphael T. Malikian. Email: rtmalikian@gmail.com. GitHub: github.com/rtmalikian

**ORCID:** 0009-0008-4326-4205

**Article type:** Retrospective observational study with reflective practice analysis

**Keywords:** large language models; AI agents; tool use; model context protocol; open-source software; pull requests; software reliability; clinical AI infrastructure

**Author background, in brief.** The author trained in medicine in the United Kingdom and completed family medicine residency training in the United States. He later bootstrapped a direct primary care clinic and ran it as chief executive. He now builds clinical artificial intelligence tools and contributes code to open-source health and AI projects. His earlier peer-reviewed work includes a four-station knee simulator study published in 2013 [18]. This paper covers the layer underneath the agent: the memory stores, tool bridges, approval screens, and crawler