Install any skill in seconds. Free to start, no credit card required.
Get Started Free →chorus-aidlc/plugin-maintenance
Guide for modifying the Chorus plugin (Claude Code, Codex, OpenClaw, Kiro, and Pi ports), updating skill documentation, and releasing new plugin versions.
Passed 21 of 23 test cases with the skill, and 4 of 23 without it.
and 23 more in the full report
17 fixed4 passed either way2 still failing0 broke2 errored
Tokens, turns and time describe resource use, not a dollar price.
This benchmark measured an earlier version of the skill.
A newer attempt on gemini-3.8-flash on 10/19/2026 did not produce a publishable comparison (not comparable). The snapshot above is the previously measured run.
One of 17 test cases the skill fixed. Same model, same prompt, with and without the skill.
Where is the version string located in OpenClaw's plugin manifest openclaw.plugin.json?
| Checked | Without skill | With skill |
|---|---|---|
| Identifies that openclaw.plugin.json contains no version field and versioning is specified in package.json. | Missed | Passed |
The instructions your model reads when the skill loads.
How to modify the Chorus plugin, update skill documentation, and release new versions. Seven plugin packages are maintained in parallel — Claude Code, Codex, OpenClaw, Kiro, Pi, dsh (DeepSeek Harness), and Hermes Agent — plus the standalone skill surface. That is eight skill surfaces total; when you change skill content, sweep all eight (see Skill Content Changes — Seven Surfaces, which also lists the Hermes copy). The Hermes port has its own section: Hermes Agent port.
TRAE CLI is not an eighth package. TRAE CLI (traecli, a Codex fork) installs the Codex package plugins/chorus/ unchanged: it reads .agents/plugins/marketplace.json and .codex-plugin/plugin.json. TRAE has no separate skill copy, version sequence, or installer. The one package serves two runtimes, so every Codex-package change must keep working under both. See Codex/TRAE plugin hooks.
.claude-plugin/
marketplace.json ← Claude Code marketplace registry (version here)
public/chorus-plugin/ ← Claude Code plugin package
.claude-plugin/
plugin.json ← Plugin metadata (version here)
hooks.json ← Hook definitions (SubagentStart, etc.)
bin/ ← Hook scripts (bash) — stateful via API state-get/set
skills/
chorus/SKILL.md ← Core skill
develop/ idea/ proposal/ quick-dev/ review/ yolo/SKILL.md
agents/ ← Reviewer agents as .md (Claude Code style)
proposal-reviewer.md
task-reviewer.md
plugins/chorus/ ← Codex plugin package (separate from Claude Code); also installed by TRAE CLI
.codex-plugin/
plugin.json ← Plugin metadata (version here; shared by Codex + TRAE)
hooks.json ← NEVER change the command strings (forces users to re-trust hooks)
hooks/ ← Hook scripts (bash) — intentionally stateless, shared by both runtimes
on-session-start.sh
on-post-submit-proposal.sh
on-post-submit-for-verify.sh
on-post-verify-task.sh
hook-runtime.sh ← Codex-vs-TRAE runtime detection + tool_response UUID parsing
chorus-mcp-call.sh ← MCP helper (has hardcoded clientInfo version)
hook-output.sh
skills/
chorus/SKILL.md ← Codex port — mentions $skill syntax, ~/.codex/config.toml
develop/ idea/ proposal/ quick-dev/ review/ yolo/SKILL.md
chorus-proposal-reviewer/SKILL.md ← Reviewers as skills in Codex (no agents/)
test cases passed, vs 4 / 23 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.