Claude Acts as Sole Engineer to Ship GitHub Action in a Week
Fit_Engineer9548 · reddit · 2026-07-15
A non-technical founder shares how they used Claude as their "sole engineer" to build an open-source GitHub Action and publish it to the Marketplace in just one week.
The Product: IntentGuard
- A GitHub Action that checks whether AI-generated PRs actually align with Linear tasks.
- Outputs ALIGNED / MISALIGNED / UNCLEAR, essentially using AI to review AI's homework.
Claude's Actual Contributions
- Set up a README-first repository structure and prioritized an offline mock mode.
- Implemented GitHub / Linear connectors and the Claude decision engine.
- Wrote 106 tests, all runnable without an API key.
- Created the repository, generated demo GIFs, published the Marketplace listing, and even curated good-first-issues.
Security Design
- Treated PR text as attacker input.
- Avoided checking out code and using pullrequesttarget.
- Handled prompt injection via delimiters and secret redaction.
The author was most surprised that Claude implemented many security measures autonomously after "PR content is untrusted input" was specified, without needing explicit reminders.
More from coding & agent
- HeyGen adds a media-sourcing skill for coding agents with 75k images and 10k tracks — HeyGen · 2026-07-22
- Agent search bottlenecks are now about variance, not raw latency — rohanpaul_ai · 2026-07-22
- LangSmith adds tracing for Pipecat, LiveKit, OpenAI Realtime, and Gemini Live — LangChain · 2026-07-22
- An MCP server signs every AI agent tool call into a verifiable Merkle chain — Funky_Chicken_22 · 2026-07-22
- Annotated transcript of a Claude Code team interview is now available — trq212 · 2026-07-22
- Claude Code skill uses 10 Markdown rules to make outputs ADHD-friendly — alex_verem · 2026-07-22