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 skills/devstride/claude-plugin/ultracode-buildnpx skills add devstride/claude-plugin --skill ultracode-buildgit clone --depth 1 https://github.com/devstride/claude-pluginWhat 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 | $0.00030 | $0.03869 |
| Opus 5 | $0.00015 | $0.01935 |
| Sonnet 5 | $0.00006 | $0.00774 |
| Haiku 4.5 | $0.00003 | $0.00387 |
Grade A, and why
ultracode-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 2d 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 — 249 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Human output. Read ${CLAUDE_PLUGIN_ROOT}/skills/build-item/references/plain-language-output.md once per top-level run; composed skills reuse it. Apply it to every message.
The build engine for one scoped story: understand → build → verify → risk check → hand back.
Invoked by build-item once the branch exists and the item is In Progress.
Argument — item number + one-line goal, optionally followed by profile: <name> and
review-moment: release-deferred|pr-boundary (e.g. I20130 enforce the seat-count invariant profile: standard review-moment: release-deferred): $ARGUMENTS
Config: .claude/ds-config.json (verify.*, baseBranch, generated.*,
preCommitWiringChecks, conventionsDoc, lessonsDoc, review.localAssistCommand, profile, profileOverrides) —
authoritative over any literal here.
Coding conventions live in conventionsDoc (AGENTS.md), not the config. lessonsDoc (inline
fallback .claude/ds-lessons.md) is the repo's lessons store, distilled from past review
findings — this skill READS it and never writes it; review owns every write. An
absent or empty lessons file is a valid state: proceed exactly as you would without it, no
note, no prompt, no setup step.
Engineering economy. Read
${CLAUDE_PLUGIN_ROOT}/skills/ultracode-build/references/engineering-economy.md before choosing
an approach or launching agents. It governs reuse, DRY/YAGNI, bounded parallelism, task-sized
Claude models/effort, and the optional read-only local support call.
Delivery profile. How much of this engine's budget a story gets is set by the delivery
profile — the canonical contract is
${CLAUDE_PLUGIN_ROOT}/skills/plan/references/delivery-profiles.md, and this skill honours
three of its knobs: understandReaders (phase 1), fixFloor (phase 3), and storyVerify
(phases 2 and 4). Resolve it ONCE, before phase 1,
and announce it with its source:
- Passed in the invocation (
profile: <name>— thebuild-itempath): use it as given, do not re-resolve, and announce it as such ("profile: standard — from the invocation"); the caller already walked the order and reported the underlying source. The name must be one of the three the contract defines; anything else (profile: standart) is a stop-and-ask, never a guess or a silent fall-through — with an unknown name no column supplies the knobs below. - Standalone (no profile in
$ARGUMENTS): walk the contract's resolution order — its root marker needsget_item(view: 'full'), since the default projection omits the description and a summary read silently falls through to the config default. Announce the result and where it came from ("profile: prototype — from the plan root", "… — from config", "… — default").
Then apply profileOverrides to those knobs within the contract's floors; report and ignore an
unknown knob. The immediate auth/migration/deployed-contract verifier below cannot be overridden
away.
Stop and ask ONLY at a genuine fork: an ambiguous or risky review finding, scope that turns out human- or infra-gated, or a destructive/outward-facing action.
1. UNDERSTAND
Build an evidence-backed picture of "done" before writing code. Provision readers proportional
to the story — don't reflexively fan all six. The profile's understandReaders is the CAP on
that fan-out; the triage below still decides how much of the cap a given story uses.
Load lessonsDoc here, alongside conventionsDoc. It is small and capped per
${CLAUDE_PLUGIN_ROOT}/skills/review/references/lessons-format.md, so read it inline — it is never a
reason to spin up a reader. Lessons are ADVISORY hypotheses, NOT rules — this is the difference between them and
conventionsDoc, which is human-owned and binding. A lesson is machine-distilled from past
findings and never independently reviewed, so it can be stale, overgeneralized, or simply
wrong. Before letting one shape the implementation, check that it actually applies to THIS
story's code; where it does, handling it up front is the cheapest place to deal with a known
mistake class. Where it does not, ignore it — silently, with no note. A lesson never overrides
conventionsDoc, the item's spec, or your own reading of the code. Absent or empty file →
skip this paragraph entirely and build as usual.
What ships with it
3 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 2d ago First seen · 249 lines · 30 tokens per session scan A 0770525a9229
ultracode-build is a skill published in the GitHub repository devstride/claude-plugin (2 stars, last pushed 3d ago), licensed MIT. It adds 30 tokens to every session and 3,869 once invoked, about $0.0002 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 skills, from other repositories
systematic-debugging
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.
brainstorming
You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.
chat-pet-sprite-creation
Use when creating or changing VS Code chat pet sprite art, sprite sheets, state animations, eye treatments, Stable/Insiders variants, or pet transitions under src/vs/workbench/contrib/chat/browser/widget/media/chatPet.
cpu-profile-analysis
Analyze V8/Chrome CPU profiles (.cpuprofile) and DevTools trace files (Trace-.json). Use when: profiling performance, investigating slow functions, comparing code paths, finding bottlenecks, analyzing timeToRequest, understanding call trees from sampling profiler data, analyzing layout/paint/rendering, investigating…
agent-host-chat-contributions
Build and review cross-cutting agent-host chat behavior through lifecycle contributions. Use when adding turn lifecycle side effects, prompt or context injection, restored-history transformation, protocol-action observation, or when reviewing changes that add code to AgentSideEffects or AgentService.
auto-perf-optimize
Run agent-driven VS Code performance or memory investigations. Use when asked to launch Code OSS, automate a VS Code scenario, run the Chat memory smoke runner, capture renderer heap snapshots, take workflow screenshots, compare run summaries, or drive a repeatable scenario before heap-snapshot analysis.