How to use Claude to build a shell recording app
ShellSight's author used Claude Code and Opus 4.5 to build shell recording and replay for SSH and Kubernetes, revise storage for scale and multi-tenancy, and fix security findings after manual verification.
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Build [shell recording application] for [SSH and Kubernetes environments]. Implement recording, upload to [object storage], and browser replay. Support [SSO requirements] and [multi-tenancy needs]. Revise the design as needed, then fix security findings I have manually verified.
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source poster / publisher
- takladev
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- The author reports building ShellSight for SSH and Kubernetes session recording and replay, spending roughly $50 on Claude and using it to fix manually verified security findings.
- Human involvement
- The author changed storage requirements, ran security scans and manually verified findings before asking Claude to fix them.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- Use Claude Code and Opus 4.5 to develop the ShellSight application.
- Reverse-engineer scriptreplay code to build a React console player.
- Implement shell recording and S3 upload for SSH and Kubernetes access.
ShellSight's author reports developing the application entirely with Anthropic Opus 4.5 through the Claude Code CLI, spending roughly $50 on Claude. The application monitors, records and replays shell access through SSH and Kubernetes pod debugging. Claude reverse-engineered scriptreplay code to create a React console player and handled SSO integration.
For SSH sessions, the application changes configuration so that login displays a banner and starts a terminal recorder. When the shell exits, the recording goes to an S3 bucket for console replay. Kubernetes access uses ephemeral containers to start recording and upload the file on exit. Admission controls deny kubectl exec access and prevent use of other ephemeral containers.
The author used the project to examine whether AI could handle realistic development beyond an initial implementation. They changed the storage model multiple times for scale and multi-tenancy. They also ran security scans, manually verified the findings and had Claude fix the confirmed issues.
The post says the chosen components allow on-premises deployment, including air-gapped environments. It does not report a production rollout or an independent security assessment.
Who it's for: Developers evaluating agent-assisted implementation and revision of operational software, with direct responsibility for reviewing security findings.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Hacker News · takladev
Frequently asked questions
How were security findings reviewed?
The author ran security scans on the generated code, manually verified the findings and then used Claude to fix those findings.
How does ShellSight handle Kubernetes shell access?
It records access through ephemeral containers and uploads recordings on exit. Admission controls deny kubectl exec and ephemeral containers other than those provided.
More details
Remaining steps
- Change the storage model repeatedly for scale and multi-tenancy.
- Run security scans and manually verify the reported findings.
- Have Claude fix the verified security findings.
About this account
The author attributes application development to AI, including replay and SSO work, while directing revisions and security remediation.
Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=46530971. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: takladev on Hacker News
Source published
Added to Agents In Use
Source checked
Tools and topics
Claude Code CLI · Opus 4.5 · React · scriptreplay · S3 · Kubernetes
SSH · Kubernetes · Session recording · Security review
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.