implementlisted
Install: claude install-skill taipt1504/claudehut
# Implement (phase 5 of 7)
Execute the approved plan **test-first**, producing code that satisfies the spec and passes every applicable
rule. This skill is preloaded into `claudehut-implementer` (which runs in an isolated worktree) and is also
the main thread's playbook when implementing directly. The per-file tech-stack standards live in the
project's `.claude/rules/` tree and **auto-load by path** as you touch matching files — follow them; this
skill carries the workflow discipline and the deeper playbooks.
## Iron Law
```
NO PRODUCTION CODE WITHOUT A FAILING TEST FIRST
```
Wrote production code before the test? Delete it. Start over. **No exceptions** — don't keep it "as
reference," don't "adapt" it while writing the test, don't even look at it. Delete means delete.
**Violating the letter of this law is violating the spirit of it.**
## Preconditions (the write gate — tier-aware)
Production writes are denied by the `PreToolUse` gate until: `reuse_scan=true` (**every tier** — Discover
produces it), plus — **in the `full` tier only** — `spec_path` and `plan_path` set, plus — **every tier,
the skill rail** — *this skill was invoked for this task*. Invoking `claudehut:implement` opens that rail (a
`PreToolUse(Skill)` recorder hook proves the call; entering Discover/Brainstorm closes it for the next task)
— so if the gate sent you here, the rail is now open. In the
`trivial`/`small` fast lanes, reuse-scan + the skill rail open the gate **provided** the change stays
within