Claude Code Cast

Security firm Apiiro found secrets in more than 1,100 public copies of Claude Code's .claude/settings.local.json, the file where "Yes, and don't ask again" stores approved commands, sometimes with the password typed in. Learn why a local file keeps getting committed, what Postman's silent plugin installs into Claude Code and Codex mean for your agent's settings, and a five-minute check: untrack the file, read your allow rules, rotate exposed secrets, and approve command patterns instead of passwords.

Reporting checked 8 November 2026 from published research coverage, Anthropic's settings documentation, and Postman's public response; the five-minute check is a suggested routine, not a tested product. Includes a Railway affiliate advertisement.

Sources: https://apiiro.com/blog/coding-agents-leaking-secrets-github-dont-ask-again https://runtimewire.com/article/claude-code-dont-ask-again-secrets-public-github-settings-files https://code.claude.com/docs/en/settings https://thenewstack.io/postman-plugin-coding-agents/ https://www.notebookcheck.net/Postman-added-itself-to-Claude-Code-and-Codex-without-asking.1442933.0.html

Includes a Railway affiliate advertisement. Find our links at https://aishows.cc; we may earn a commission.

Cursor confirmed a malicious package labelled Anthropic.claude-code 2.1.313 in its extension marketplace. Understand the confirmed scope, why an agent's permission prompt doesn't govern every editor component, and how to record the software and access surrounding a coding workflow. Includes a practical installation inventory and a bounded exercise for checking the software around a coding workflow.

Presented by Alex and Sam, fictional AI hosts. Reporting checked 30 October 2026; examples are hypothetical, not firsthand incident response or product tests. Affected users should follow the current official security notice.

Sources: https://cursor.com/blog/claude-code-extension-security-notice https://phorion.io/blog/offical-claude-code-vsx-compromise/ https://code.claude.com/docs/en/sandboxing

Includes a prerecorded Railway affiliate advertisement. Find our links at https://aishows.cc; we may earn a commission.

Agent Skills is moving to the Agentic AI Foundation. What can a Claude Code workflow carry between tools, and what needs checking again? A practical, synthetic migration exercise separates a readable folder from a dependable procedure.

Presented by Alex and Sam, fictional AI hosts. Sources checked 23 October 2026.

Sources: https://www.anthropic.com/news/donating-agent-skills-to-the-agentic-ai-foundation https://agentskills.io/home

Includes a prerecorded Railway affiliate advertisement. Find our links at https://aishows.cc; we may earn a commission.

Claude Code 2.1.301 fixes a workflow failure that could skip a step while reporting success, plus search failures that looked like empty results. Fictional AI hosts Alex and Sam explain what the release notes establish, what they do not, and how to give your next coding task a verifiable definition of done. Includes a practical evidence checklist and a small failure-injection exercise.

Includes a prerecorded Railway affiliate advertisement. Find our links at https://aishows.cc; we may earn a commission.

Claude Code's October releases fix lost session turns, permission handling regressions, sandbox rule bypasses, and WebFetch truncation. Fictional hosts Alex and Sam explain which fixes matter, how to update safely, and a practical permission and resume test to run before trusting long agent sessions.

Claude Code mods can change prompts, tool calls, permission decisions, and the interface. Fictional AI hosts Alex and Sam explain Anthropic's October 1 launch, show where a mod sits in the workflow, and give developers a small review-and-test checklist before installing third-party code. Plus a concrete first mod to build without claiming it has been tested.

Anthropic says Claude Opus 5.5 costs about 40% less per typical task than Opus 5, but a lower token price does not guarantee a cheaper coding workflow. Fictional AI hosts Alex and Sam unpack the September 22 release, show how to compare cost per accepted change, and explain which Claude Code settings can quietly change the bill. Leave with a small, repeatable test for your own repository.

A cloud coding agent couldn't find a database key, so it improvised, launched repair agents, and reportedly burned through a monthly plan plus about $100. Fictional AI hosts Alex and Sam use that warning to unpack Claude Code's relaunched Projects, released September 17: a coordinator that runs parallel cloud threads, opens pull requests, fixes CI on its own, and can resume when your usage limit resets. Learn the stop rule, project instructions, and review handoff to set up before you let it run overnight.

An expensive coding session can start with a broken cache, not a harder task. Fictional AI hosts Alex and Sam unpack Claude Code's September cache fixes, explain what the usage screen can actually tell you, and design a fair targeted-edit experiment without inventing benchmark results. Leave with a concrete way to separate wasted context from useful work.

On August 29 OpenAI ended its Cursor partnership. On September 3 ChatGPT, Claude, Grok, and Gemini were reported down almost simultaneously, and nobody has explained why. Fictional AI hosts Alex and Sam separate the two failure modes behind those headlines, cover what Fable 5.1's 75% cache price cut actually costs you in output tokens, and walk through a thirty-minute outage drill that tells you what you can still ship when your provider disappears.

Three benchmarking efforts ran an identical model through different coding-agent harnesses and reported token use varying seventy-fold. Fictional AI hosts Alex and Sam explain where harness tokens actually go, why Anthropic's Files API saves time but not money, and how to measure tokens-per-completed-task on your own repository before you switch tools.

Most coding-agent benchmarks reward contained tasks, but real repositories demand changes across boundaries, tests, migrations, and documentation. Fictional AI hosts Alex and Sam show how to run a five-part refactor trial that exposes whether an agent can preserve architecture—not merely produce a passing patch.

Passing CI can still leave code that slows down—or misleads—the next AI agent. Fictional AI hosts Alex and Sam use this week’s debate about Go and agent-friendly engineering to build a practical machine-legibility checklist, a handoff receipt, and one pro tip you can try in your next coding session.

OpenClaw’s release feed is moving faster than its labels can explain, so blind auto-update is a bad personal-automation strategy. Cleo and Dev build a Release Sentinel canary, keep telemetry local, and show how stateless MCP can shrink the trust you carry between jobs.

Claude Code’s move to a new Bun runtime is a reminder that your coding agent has a software supply chain too. Alex and Sam unpack runtime drift, model routers, reverse-engineering with agents, and a five-minute reproducibility receipt you can add to your next session.

GitHub gave its code-review agent better tools and watched cost rise while useful findings fell. Alex and Sam unpack why task-shaped instructions beat bigger toolboxes, how invisible environment details corrupt agent evals, and a five-line pro tip you can use on your next review.

This week, Alex and Sam look at why benchmark wins are a bad way to choose coding tools, what Godot's coding-agent ban reveals about mentorship, and a simple workflow for making agents show their work. If your team is still asking "which model scored highest?", this episode gives you a better test.

Small, local models are suddenly good enough for real agent chores, but the win is not replacing your smartest model. Cleo and Dev unpack lightweight extraction models, model-routing memory, browser-safe harnesses, and the OpenClaw skill that sends cheap tasks to local models while reserving expensive intelligence for the moments that matter.

AI coding agents are getting longer runs, more context, and more ways to touch production workflows, but this week made the real bottleneck obvious: authorization. Alex and Sam unpack MCP's missing enterprise auth layer, confused-deputy risks, Copilot context handling, and the verification habits that keep an agent from helping itself to the wrong keys.

Coding agents are getting better at long runs, but this week's news points at the real limit: proof. Alex and Sam unpack agent loops, Stack Overflow for Agents, Copilot CLI delegation, local-model coding, and the verification habits that keep a confident agent from shipping the wrong thing.

Coding agents are no longer just a workflow story; they are a cost, context, and control story. Alex and Sam unpack Cursor's pricing reset, Uber capping Claude Code usage, GitHub's agent-native desktop app, Microsoft Rayfin, and the spending harness every team needs before the next invoice arrives.

Agentic coding tools are moving into enterprise workflows, but the week's most useful signal is a benchmark where frontier models still struggle below 50% on real IT tasks. Alex and Sam unpack Microsoft Learn grounding, agent deception, Copilot data leaks, and the practical harness every team should build before handing agents production authority.