Borrowing it
Nothing to install: this file belongs to Black-coffe/vulyk. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/Black-coffe/vulyk/main/.claude/commands/vulyk-plan.mdgit clone --depth 1 https://github.com/Black-coffe/vulykWrote 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/black-coffe/vulyk/vulyk-plan)<a href="https://agentmods.dev/commands/black-coffe/vulyk/vulyk-plan"><img src="https://agentmods.dev/badge/commands/black-coffe/vulyk/vulyk-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 5.1 | $0.00017 | $0.01154 |
| Opus 5 | $0.00009 | $0.00577 |
| Sonnet 5 | $0.00003 | $0.00231 |
| Haiku 4.5 | $0.00002 | $0.00115 |
Grade A, and why
vulyk-plan scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured today.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 17 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Enter Queen mode for: "$ARGUMENTS"
- Classify the tier per the routing matrix in CLAUDE.md. Announce it. Tier 0-1: skip ceremony - say so and either do it (Tier 0) or write a single story and stop for approval (Tier 1). The ceremony floor: brief.md and
## Requirementsquotes exist at Tier 2+; trace-check runs whenever stories exist; Tier 0-1 gets none of it. - Brief. Write
docs/specs/<slug>/brief.md: the request VERBATIM as a>blockquote - the user's words, not your restatement - piped throughbash scripts/redact.sh(briefs are committed; secrets are not), with the date. Everything downstream traces back to this file; a paraphrase here poisons every gate built on it. A bug report is a spec too: the error text, the stack trace and the reproduction go in verbatim, and the ask is the observable that must hold afterwards. - Briefing questions. Before asking the human anything, look it up - the repo,
memory/, the brief itself answer most questions. Ask only what is irreversible, costly, a vendor choice, or a business rule no code can reveal. One question at a time, each with a recommended default so silence has a safe meaning. Append answers to brief.md under## Answers(also verbatim). Then confirm the spec (stage 01). Show the brief back - the quote and the answers, nothing else - and at Tier 3-4 stop for one word before spending recon: every gate downstream answers to this text, and a wrong word here is built correctly by everything below it. At Tier 2 show it and continue unless corrected; silence means the text stands. - Recon, not reading. Check
memory/memory.mdand relevantmemory/map/slices first. Dispatchdrone-scout(parallel, up to 4) only for territory the map does not cover or marks stale. You do not open source files yourself. - Plan. For Tier 2: draft the plan inline. For Tier 3-4: delegate synthesis to
queen-plannerwith the brief, scout reports and map pointers attached; for Tier 4 also request alead-architectconsult on the central design fork before stories are finalized. Both dispatches carrymodel: <TOP_MODEL>- the alias the session brief announced (fablewhere the plan carries Fable 5.1,opusotherwise;bash scripts/top-model.shif the brief scrolled away). The agent files sayopusas the floor; the per-invocation parameter is where the plan-aware upgrade happens, and forgetting it plans on the floor. Plan file followstemplates/plan.md- contracts between concurrent stories are decided here, in the one context that has seen the whole plan. - Stories. Ensure
docs/specs/<slug>/contains plan.md plus one story file per unit of work (template:templates/story.md). Each story: verbatim## Requirementsquotes from brief.md, files to touch, acceptance criteria, verification command, map slice pointer, model tier of its worker, and itswave:/blocked_by:lines. Wave rule: stories in one wave run concurrently, so their## Filesmust be disjoint - when two stories need the same file, either merge them, move the shared file into its own earlier story, or push one to a later wave viablocked_by. - Check the stories, deterministically. Run
bash scripts/wave-check.sh docs/specs/<slug>andbash scripts/trace-check.sh docs/specs/<slug>- both deterministic, both free. Wave-check catches the collision that would silently eat a worker's diff mid-build, the declared path that does not exist in the tree, and the story whose verification command cannot turn red for its own files; trace-check catches the story nobody asked for and the requirement no story carries. Fix findings in the story files now; a plan approved with either defect ships it. - Coverage, from outside the plan. Dispatch
drone-coveragewith exactly two paths: the brief and plan.md. It never sees the story files - that is the whole mechanism, because a check that reads the planner's own output cannot be independent of it. Its report is evidence, not authority: an absent or partial ask is either a story you owe or a line the human will call context. Do not resolve that yourself. - Stop for approval. Present: tier, story list with one-line summaries grouped by wave, open assumptions (Law 1), estimated token posture (how many worker dispatches, where top-model tokens will be spent), and all three verdicts - wave-check, trace-check and coverage - including which brief lines no story covers, because deciding "that line is context, not a requirement" belongs to the human, not the planner. Do NOT proceed to implementation - that is /vulyk-build, run after human approval. When it comes, replace the
**Approved:**placeholder in plan.md with the owner and the date - that line is stage 02's confirmation, and it is the one/vulyk-buildreads.
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.
- today Changed 47fca013c74b
- yesterday Changed 03329351a390
- 6d ago First seen · 17 lines · 17 tokens per session scan A ad61343be9cd
vulyk-plan is a command published in the GitHub repository Black-coffe/vulyk (2 stars, last pushed yesterday), licensed MIT. It adds 17 tokens to every session and 1,154 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
oss-docs
(user-owned) oss-docs harness — README 4-locale + docs-site publishing for the moai-adk-go OSS project (adk.mo.ai.kr, Vercel auto-deploy).
review
Code review with security and @MX tag compliance check.
fix
Auto-detect and fix LSP errors, linting issues, and type errors.
gate
Run lint+format+type-check+test in parallel (pre-commit quality gate).
project
Generate project documentation (product.md, structure.md, tech.md, codemaps/).
audit-agents-skills
Audit quality of agents, skills, and commands in a Claude Code project.