Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/vsladkov/claudex-stereo/plangit clone --depth 1 https://github.com/vsladkov/claudex-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/plan)<a href="https://agentmods.dev/commands/vsladkov/claudex-stereo/plan"><img src="https://agentmods.dev/badge/commands/vsladkov/claudex-stereo/plan.svg" alt="Measured on agentmods" 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 | $0.00014 | $0.04987 |
| Opus 5 | $0.00007 | $0.02493 |
| Sonnet 5 | $0.00003 | $0.00997 |
| Haiku 4.5 | $0.00001 | $0.00499 |
Grade A, and why
plan 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 4d 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 — 349 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 and apply its routing, foreground
agent, validation, persistence, quoting, and background-job rules. The rules below are
step-specific.
Run the planning phase of the Stereo workflow. The current Claude session orchestrates; the selected planner and plan-reviewer perform their routed steps. Never implement from this command.
Raw slash-command arguments:
$ARGUMENTS
Arguments and modes
After reading the routing skill, parse every argument before inspecting the repository or starting a routed step:
--planner <model>selects the drafter and defaults toclaude:fable.--planner-effort <none|minimal|low|medium|high|xhigh|max>overrides effort for a Codex-routed planner.--plan-reviewer <model>selects the plan reviewer and defaults tocodex:sol.--plan-reviewer-effort <none|minimal|low|medium|high|xhigh|max>overrides effort for a Codex-routed plan reviewer.--effort <none|minimal|low|medium|high|xhigh|max>is the command-wide default for Codex-routed roles that have no role effort flag. When no active role is Codex-routed, accept--effortbut report that it is inert rather than silently dropping it.--max-plan-rounds <n>defaults to 6.--slot <name>selects the durable plan slot this run stores into and defaults todefault. Slot names are trimmed, lowercased, may contain only letters, digits, hyphens, and underscores, and must start with a letter or digit. Relay the CLI's validation error verbatim.--draft-onlyruns one draft step, stores it, and stops.--review-onlyreviews the stored plan exactly once and stops.--plan-file <path>reviews that external plan exactly once and is valid only with--review-only.- Without a mode flag, run the complete draft-plus-review/revision phase.
Reject missing values, duplicate role or role-effort flags, invalid effort or round values,
unknown flags, unknown claude:* values, and both mode flags together. Accept claude:inherit
alongside claude:session and the four explicit Claude aliases. Accept a Codex selection with or
without the codex: prefix and reject codex:claude:*. The removed --planner-model and reviewer
--model flags are unknown; report the role-named replacements.
For --review-only, reject task text, --planner, and --planner-effort. Reject --plan-file
with --draft-only or the full phase and say that it requires --review-only. For --draft-only,
reject --plan-reviewer, --plan-reviewer-effort, and --max-plan-rounds. Reject a role effort
flag when its selected role is Claude-routed. Resolve each active Codex role through the routing
skill's role effort > command-wide effort > workspace role effort > model default hierarchy. Full
phase and --draft-only require task text; if it is empty, ask the user what to plan.
Define these invocation placeholders before any routed step:
<plannerSelectionArgs>=--model <effectivePlannerModel> <plannerEffortArg>.<reviewSelectionArgs>=--model <effectivePlanReviewerModel> <planReviewerEffortArg>.<plannerEffortArg>and<planReviewerEffortArg>are--effort <resolved effort>when the corresponding role's resolved effort is non-null, and are omitted entirely otherwise.<slotArg>=--slot <slot>when this run targets a non-default slot, and is omitted entirely for thedefaultslot.
The task command injects no server-side effort default, so omitting the planner --effort
silently loses a resolved max for a gpt-* planner. plan-review does default a missing
--effort to the selected model's pair default, so the reviewer placeholder enforces consistency
rather than correcting runtime behavior.
The codex-result-handling stop-after-findings rule applies at explicit user-decision points and
to --review-only. During the full phase's review loop, revise automatically.
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.
- 4d ago First seen · 349 lines · 14 tokens per session scan A bfa8ccbe3c57
plan is a command published in the GitHub repository vsladkov/claudex-stereo (3 stars, last pushed 25d ago), licensed Apache-2.0. It adds 14 tokens to every session and 4,987 once invoked, about $0.0001 per session on Opus 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
rest
Maude's end-of-session ritual — digest + tomorrow's starting point + close-the-loop check + full save fan-out. Run before you stop, so the next session opens cleanly. Session-end is the right time to pay network cost, so Tier 2 writes happen here when registered.
save
Save the current session digest. Fans out across every writable memory tier per the house-map. Tier 0 always; Tier 1 if reachable; Tier 2 only if registered writable with auth. Each tier independent — failures degrade gracefully.
wake
Maude's session-start ritual — brief + house-walk + 1-3 things to know first. Runs at session-start manually if SessionStart hook didn't catch you, or anytime you need a clean re-orient. Cheap by design — Tier 0 always, Tier 1 if cached-up, Tier 2 never.
notice
Maude surfaces patterns from the turn-by-turn trace — recurring topics, repeated mistakes, time-of-day patterns, sessions that keep ending in the same place.
teach
Tell Maude something about yourself directly — how you work, your rhythm, what help you want — and she records it in her cross-project profile of you. Use when you want her to KNOW a fact rather than wait for her to infer it. Tier 0, user-global, no network.
verify
Run Maude's programmatic project audit — version consistency, JSON validity, link integrity, header dates, watch-list paths, worn-framing scan. Catches "ready" before it's actually ready.