Getting it into your agent
This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.
/plugin marketplace add vsladkov/claudex-stereo/plugin install stereoWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/vsladkov/claudex-stereo/implement)<a href="https://agentmods.dev/commands/vsladkov/claudex-stereo/implement"><img src="https://agentmods.dev/badge/commands/vsladkov/claudex-stereo/implement/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/vsladkov/claudex-stereo/implement"><img src="https://agentmods.dev/badge/commands/vsladkov/claudex-stereo/implement.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00015 | $0.05782 |
| Opus 5.5 | $0.00006 | $0.02313 |
| Sonnet 5.5 | $0.00003 | $0.01156 |
| Haiku 4.5 | $0.00002 | $0.00578 |
Grade A, and why
implement scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 21d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 366 lines — stays where its author put it; the contents beside it link to each section on GitHub.
First Read ${CLAUDE_PLUGIN_ROOT}/skills/model-routing/SKILL.md, then
${CLAUDE_PLUGIN_ROOT}/skills/model-routing/pair-procedures.md and
${CLAUDE_PLUGIN_ROOT}/skills/model-routing/implement-procedures.md — plus
${CLAUDE_PLUGIN_ROOT}/skills/model-routing/worktree-procedure.md for an --isolated run or the
resume of an isolated record — and apply their rules. The rules below are step-specific; a quoted
heading names the section of a routing file that defines the step.
Run the implementation phase of the Stereo workflow. The main Claude session owns preflight, baselines, host verification, review/fix orchestration, and the final report.
Raw slash-command arguments:
$ARGUMENTS
Arguments and modes
Parse all arguments per "Selection and effort parsing" before loading state, then read the "Workspace role defaults" and say in the final effective-role note when a workspace default supplied the implementer:
--implementer <model>and--implementer-effort <effort>select the implementer and its effort.--implementation-reviewer <model>and--implementation-reviewer-effort <effort>select the implementation reviewer and its effort. Its built-in default is a Codex model: the implementation review is the last gate before commit, and the cross-ecosystem reviewer is independent of both the orchestrator and the default Claude implementer.claude:sessionis the cheaper inline choice.--max-fix-rounds <n>defaults to 4. Direct gate-fix turns and review-driven fix turns both count toward it; the implementer's own inner-loop iterations never do.--slot <name>selects the stored plan to implement and defaults todefault.--implement-onlyimplements and verifies once, then stops before review.--review-onlyreviews the current dirty/untracked implementation delta once without fixing.--base <ref>reviews the committed<ref>...HEADrange and is valid only with--review-only.--resumere-enters a recorded incomplete implementation phase after a crashed, compacted, or closed Claude session.--isolatedruns implementation in a throwaway detached git worktree and hands the delta back as a user-confirmed patch.- Without a mode flag, run the complete implement-plus-review/fix phase.
Reject positionals and both mode flags together. Reject --fresh: the implementer always starts
fresh. For --review-only, reject --implementer, --implementer-effort, and --max-fix-rounds.
For --implement-only, reject --implementation-reviewer, --implementation-reviewer-effort, and
--max-fix-rounds. Reject --base without --review-only, and reject --scope in every mode,
naming --base <ref> as the supported standalone range control: auto-detecting a default base
could silently review commits the plan never covered. Reject --isolated with --review-only and
with --base.
Reject --resume with --implement-only, --review-only, --implementer, --implementer-effort,
or --slot: the durable record owns the implementer and the plan slot, so tell the user to run
without --resume to start over. --implementation-reviewer, --implementation-reviewer-effort,
and --max-fix-rounds stay legal with it. A bare --resume against an isolated record resumes in
its recorded worktree and says so; reject --isolated --resume against a non-isolated record and
tell the user to run without --resume to start an isolated phase. On resume, the recorded
implementer — its launchArgs and userOwnedSteps — stays fixed for later fixes, while the
implementation reviewer is resolved, granted, and pinned afresh ("Pinned selections"); its recorded
selection is historical context.
Inside the full fix loop, act on findings automatically. In --review-only and at explicit
safeguard decisions, stop after presenting review findings and let the user decide what changes.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 21d ago Changed · -715 lines 4be7df3e3f00
- 1mo ago Changed · -1 lines 74802c8baf7c
- 1mo ago Changed · +1 lines da36f2d83837
- 1mo ago First seen · 1,081 lines · 15 tokens per session scan A fa906acf58e6
implement is a command published in the GitHub repository vsladkov/claudex-stereo (4 stars, last pushed 21d ago), licensed Apache-2.0. It adds 15 tokens to every session and 5,782 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
check
Quick sanity check - code (inline 6-perspective) or plan (--plan).
end
End a work session - a settlement pass that reconciles uncommitted / unpushed work + session context into git, the tracker, and a session snapshot file. --session (aliases --snapshot, --pre-compact, --compact) skips settlement and only writes the shared snapshot, keeping the session going.
note
Add a note to the active task - progress, decisions, context. --session (aliases --snapshot, --pre-compact, --compact) writes a resume-grade session snapshot without ending the session. --team writes a fact or a decision into the standing team's file.
review-round
Work through a RECEIVED round of review comments (inline markers, a review file, or PR threads) on a spec/doc/PR. Triage each item, record decisions to the tracker task, keep the artifact FROZEN, then apply all edits in ONE final pass. Use when a reviewer returns N comments to incorporate.
hub
Cross-project hub on Orca - see every project's orchestrators, ask a stopped one read-only, wake one for gated work (Orca addon).
statusline
Manage & persist statusline appearance - light/dark, compact, and which rows to show.