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-build.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-build)<a href="https://agentmods.dev/commands/black-coffe/vulyk/vulyk-build"><img src="https://agentmods.dev/badge/commands/black-coffe/vulyk/vulyk-build.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 | $0.00015 | $0.01485 |
| Opus 5 | $0.00008 | $0.00743 |
| Sonnet 5 | $0.00003 | $0.00297 |
| Haiku 4.5 | $0.00002 | $0.00148 |
Grade A, and why
vulyk-build scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured today.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 34 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Execute the approved plan: "$ARGUMENTS" (default: most recent plan in docs/specs/ marked approved).
-
Load the plan. Refuse politely if no approval marker - planning and building are separate decisions by design. Then put the build on its branch (stage 03): story commits live in their own branch, never on the default one.
git rev-parse --abbrev-ref HEAD- on the default branch (main/master, or whatever the Profile's Release / deploy row names), create and switch tovulyk/<slug>(or the naming that row prescribes) before wave 1; already on a feature branch, build there. Either way replace the**Branch:**placeholder in plan.md with the branch name and commit it with the plan. The branch is what stage 04 tests, what the owner is shown at stage 05 and what/vulyk-shipmerges; a build whose commits nobody can name later has no rollback point, andscripts/ship-check.shreads this line. -
Check the stories. Run
bash scripts/wave-check.sh docs/specs/<slug>and show its output. A collision, order violation or dangling blocker is a plan defect: fix the story files (merge, split, or re-wave) before dispatching anything. Disjoint## Fileswithin a wave is the precondition for parallel dispatch - two concurrent workers on one file silently overwrite each other. The gate runs again here, after approval, because the tree moved since planning: amissingpath is a worker about to be sent at a file that is not there, and averify-gaporno-verifystory is one whose green will mean nothing when it returns. -
Dispatch by wave, one message per wave. Launch every story of the current wave as parallel worker calls in a single message - that, and nothing else, is what makes them actually run concurrently; one call per message is a serial build wearing parallel clothes. Each story goes to its worker (
worker-code, thenworker-testwhere the story requires tests) with EXACTLY: the story file, its map slice pointer, the relevant.claude/rules/paths. Nothing more - scoped context is the law. Cap: 4 concurrent workers; a bigger wave dispatches in slices. -
Close each story as its worker returns - do not wait for the whole wave:
- Read the return report (
STATUS/FILES/TESTS/...). If a worker returned prose instead of the contract, take what it did as unverified: run the story's verification yourself via a quiet command before trusting it. bash scripts/scope-check.sh <story-file>- with per-story commits the default working-tree range is exactly this story's diff, so the numbers are finally per-story, not per-pileup.- Run the story's
## Verificationcommand (quiet variant), as many times as itsrepeat:line asks - a story that names a repeat count is telling you a single green is not evidence for this code. Red -> back to the worker path, never patched by your hands. - Commit:
git addthe story's declared files (plus the story file itself) and commit asstory(<slug>-NN): <title>. One story, one commit - it is the rollback point and the review unit. Out-of-scope files flagged by scope-check are a decision, not a default: leave them uncommitted and resolve (amend the story, or descope the change) before they ride along. - Update the story's
status:line.
- Read the return report (
-
Repair has a ceiling: two rounds per story.
NEEDS_CONTEXT-> the story was defective. Fix the story file (or answer the question), then send a fresh worker. This round counts against the plan, not the worker.WALLor red verification -> send ONE fresh worker with## Findingsattached and the repair stated as a condition to satisfy ("make X pass with Y preserved"), not as instructions to follow. Never re-dispatch the identical prompt hoping for luck.- Second failure -> stop the story: mark it
blocked, and either re-plan it or escalate the design question tolead-architect. A third identical attempt is a token bonfire. - A repair dispatch obeys the wave rule too. Before sending one alongside anything still in flight, intersect its file set with theirs exactly as wave-check does for stories. Two dispatches that land in one file minutes apart cannot be split by path afterwards: the story loses its one-commit rollback point, and nobody notices until the diff is already mixed. When the repair is a real story file, do not intersect by hand - re-run
bash scripts/wave-check.sh docs/specs/<slug>. The gate ran at step 2 against the pack as approved; a repair round changes that pack, and a check that judged a different set of stories is not a check. Same rule, same reason, as the acceptance verdict in/vulyk-review: when the pack moves, whatever judged it is re-run.
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 10d4f7f0024c
- 7d ago First seen · 34 lines · 15 tokens per session scan A ef6de36f551f
vulyk-build is a command published in the GitHub repository Black-coffe/vulyk (2 stars, last pushed 2d ago), licensed MIT. It adds 15 tokens to every session and 1,485 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
audit-agents-skills
Audit quality of agents, skills, and commands in a Claude Code project.
security-audit
Comprehensive security audit of your project AND Claude Code configuration. Analyzes secrets exposure, injection surfaces, dependencies, hook security, and produces a scored security posture assessment.
worktree
Create a git worktree for isolated feature/fix work.
init-docs
Fetch official Anthropic Claude Code docs and store as local baseline snapshot.
router-feedback
Submit feedback about router decision or model performance.
setup-pm
A command for choosing the package manager used by a project or by default on your computer. A package manager installs and manages the libraries a software project needs.