Show HN: OldHand A Claude/Codex plugin to verify the development flow end-to-end

**OldHand** is a **Show HN** project: a **Claude** / **Codex** plugin (framed as a Claude Code or Codex skill) meant to verify the development flow end to end. The author built it after relying on AI coding agents day to day for how convenient they are, then hitting a hard limit in corporate work: delivery still has to track a **Jira** or **Asana** ticket, and agent-written output often fails that bar.

Technically, the problem OldHand targets is not “can the model write code,” but whether the agent path stays aligned with an existing ticket-driven workflow. In the author’s account, agents invent or expand code that is not needed, including new logic where a reusable function already exists in the codebase. The plugin’s role is to check that flow—from ticket-shaped intent through agent work—so hallucinations and needless duplication surface before the work is treated as done.

For engineers and builders, that gap is familiar: agents accelerate implementation, but ticket systems define acceptance, ownership, and scope. Hallucinated APIs, duplicated helpers, and “helpful” extras that the ticket never asked for create review load, drift from the ticket, and reuse debt. A verifier that sits on the Claude/Codex side is aimed at the last mile—keeping agent output accountable to the same task tracker humans already use.

In market context, this sits with **Claude** and **Codex** as coding agents plus everyday PM tools (**Jira**, **Asana**). Show HN is full of agent wrappers; OldHand’s angle is narrower: not another “write the PR” demo, but end-to-end verification of the corporate development flow where ticket fidelity and code reuse matter more than raw generation speed. The source framing also ties it to the **Claude** / **Codex** / **Fable** HN discussion space rather than a full product launch with benchmarks or pricing.

Practical takeaway: treat OldHand as a workflow guardrail for ticket-bound agent coding—especially where reuse and hallucination have already bitten you—not as a general code generator. What to watch next is how far the plugin actually binds agent steps to ticket requirements and existing functions, and whether teams can plug it into real Jira/Asana loops without inventing a second process beside the one they already follow.