By Pratiyush
Spec-driven development with a Verified Traceability Graph — every requirement bound to a passing check. EARS specs, evidence-bound done, drift detection, graph-derived PRs.
Score each requirement's complexity 1–10 and recommend a subtask breakdown, then expand the over-scoped ones into bound sub-criteria. Use after a spec is drafted and before implementation, to catch requirements that are too big to prove as a single unit.
Turn approved requirements into a thin high-level design and the Architecture Decision Records (ADRs) that justify the non-obvious choices. Use after requirements are clarified and before implementation, or when a change makes a load-bearing architectural decision.
Resolve ambiguity in a raw intake or draft spec by asking the user ≤5 high-value questions, each with a recommended default. Use before finalizing a full-spec route, or when `spec.onVague` is "clarify", or whenever a requirement is under-specified.
The completion ritual — run when a feature's tasks are done and it's time to land the branch. Fresh-evidence entry gate, a fixed option menu, typed confirmation for destructive paths, provenance-checked cleanup, every step journaled. Use when the user says finish/land/merge/wrap up, or after dev-spec-kit pr is ready.
Build, refresh, and read dev-spec-kit's Verified Traceability Graph — provider-agnostic. Keep it fresh with the build wrapper, read blast radius / implements-edges / drift to answer "what breaks if this changes?", and run the staleness re-prove dance. For choosing and configuring the code-graph engine itself, see the dev-spec-kit-revitify and dev-spec-kit-graphify skills.
Executes bash commands
Hook triggers when Bash tool is used
Modifies files
Hook triggers on file write and edit operations
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Spec-driven development with a Verified Traceability Graph — every requirement bound to a passing check.
dev-spec-kit is a spec-driven development tool you drive from Claude Code (Spec-Kit-compatible layouts).
Built for Claude Code first; other assistants later.
It turns a request into EARS acceptance criteria with @check bindings, derives evidence-bound
tasks, and — unlike every other SDD tool — proves the requirement → code → test → PR edges with
real executed checks. A task physically cannot be marked done while a bound check is failing or
unrun. Proofs carry the tested tree-hash, so committing tested code keeps them green while
changed code goes stale — drift you can see and fix with one command.
Built on graphify (the code knowledge graph) — dev-spec-kit overlays the proven spec/test/PR edges and their traffic-light states on top.
Status: v0 — fully usable from source; npm publish as dev-spec-kit is deferred until the
app-dogfood phase completes (see .dev-spec-kit/DEFER.md).
dev-spec-kit route "add idle session expiry" # front door: research | quick | full-spec (+ security floor)
# write .dev-spec-kit/specs/sessions.md — EARS + @check bindings
dev-spec-kit spec tasks # spec → evidence-bound tasks (re-derive syncs bindings)
dev-spec-kit task start R-AUTH-03
dev-spec-kit check run R-AUTH-03 "SessionTest#idle" --stack java-maven --expect-red # TDD red, journaled
# implement…
dev-spec-kit check run R-AUTH-03 "SessionTest#idle" --stack java-maven # green proof @ tree-hash
dev-spec-kit task done R-AUTH-03 # THE GATE — refuses without green proofs
dev-spec-kit verify --stamp --advance # prove EVERY bound criterion in ONE suite run, then
# auto-advance fully-proven tasks (the fast path)
dev-spec-kit spec lint # static drift: orphaned @check refs (also a pre-commit gate)
dev-spec-kit graph build # Verified Traceability Graph + boards (exit 1 on drift)
dev-spec-kit drift # re-verify anything stale/red in one move
dev-spec-kit approve R-AUTH-03 --note "ship it" # human gate as a signed artifact
dev-spec-kit pr --title "Session expiry" # graph-derived PR body (+ blast radius); guard blocks non-green PRs
dev-spec-kit unlock = journaled,
time-boxed escape hatch) · opt-in DENY→FORCE→ALLOW investigative gate · PreCompact auto-saves
RESUME.md.dev-spec-kit verify --stamp proves EVERY bound criterion from ONE suite run
(no per-criterion cold starts); --advance auto-advances fully-proven tasks. dev-spec-kit spec lint
statically catches orphaned @check refs (renamed/moved tests) — folded into dev-spec-kit doctor, a
pre-commit gate, and a Stop hook. dev-spec-kit spec draft-tests scaffolds a failing, bound stub per
unbound criterion. dev-spec-kit graph build flags circular dependsOn chains.judge kind: an LLM verdict for the genuinely-unmeasurable (tone, error-copy
actionability, a transcript vs a rubric) — recorded as a second-class proof (rendered
⚖️ judged, never an executed green) and blocked on full obligations by default. Default
harness mode is free (the Claude Code agent supplies the verdict, no API key); optional
api mode calls a small Anthropic model headlessly..dev-spec-kit/config.json — TDD, flaky retries,
kind-aware runners (unit/integration/api/e2e/visual/parity/judge), app spin-up lifecycle for e2e,
custom stacks as pure config, gate packs (security/contracts/nfr/rollback) with trigger words
that floor risky requests to full-spec..dev-spec-kit/laws.md + scoped laws/*.md): always-on / file-match / summon-on-demand rules,
personal→project inheritance, #[[file:…]] doc injection — dev-spec-kit laws shows what's in force.dev-spec-kit log) · generated LEDGER.md/TRACKING.md/RESUME.md boards that cannot lie ·
warn-on-repeat from the learnings ledger · dev-spec-kit dashboard (emoji cockpit with completion %,
traffic lights, drift banner, the code graph, and every .dev-spec-kit/*.md rendered readable).dev-spec-kit wave plan/start/done — no-shared-files waves, fetch-first worktrees
branched from origin's current tip (stale-base disasters designed out), merge-gated
provenance-checked cleanup.LLM-powered knowledge base from Claude Code and Codex CLI sessions. Follows Karpathy's LLM Wiki pattern.
npx claudepluginhub pratiyush/dev-spec-kitFeature development with code-architect/explorer/reviewer agents, CLAUDE.md audit and session learnings, and Agent Skills creation with eval benchmarking from Anthropic.
Production-grade engineering skills for AI coding agents — covering the full software development lifecycle from spec to ship.
Comprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
Access thousands of AI prompts and skills directly in your AI coding assistant. Search prompts, discover skills, save your own, and improve prompts with AI.
Lazy senior dev mode. Forces the simplest, shortest solution that actually works: YAGNI, stdlib first, no unrequested abstractions.
Complete developer toolkit for Claude Code