legacy-coderlisted
Install: claude install-skill vraj-ai/skills
# /legacy-coder — Former implementation chain
This is the preserved legacy workflow. The active autonomous pipeline is
`goals`; do not auto-trigger this skill for ordinary implementation requests.
Orchestrate the build-and-hand-off loop: figure out what to work on from the
docs, implement it test-first against a machine-checkable gate, self-check the
obvious corners, then pass it down the line. This is the counterpart to `/legacy-planner`
(which produces the tickets — and the invariants — this skill consumes).
## Your place in the fleet loop
The fleet is a pipeline on the project's board, one skill per stage, each stage
run by a **different model** so that no stage ever reviews its own work:
```
Planned → Agent Ready → Coding → Debugger Ready → Debugging → Review Ready → Reviewing → Done
└ /legacy-planner ─┘ └──── /legacy-coder ────┘ └──── /legacy-debugger ────┘ └──── /legacy-reviewer ────┘
```
### Stage-parent session
The default coder is Kimi K3 in a separate top-level Pi session through OpenRouter,
with high reasoning. It may use Kimi K2.7 Code helpers only when Kimi K3 is the
independent stage parent. If `/legacy-coder` was launched as a native child of another
session, it must work directly and cannot create nested children. The GitHub
Project item, GitHub issue, git/PR artifacts, and handoff bridge this session to the
other stage parents.
**Your board moves:** claim from **`Agent Ready`** → **`Coding`** before writing
any code → **`Debugger Ready`** when