By kborovik
Compressed spec-driven development. Seven commands (/sdd:design, /sdd:spec, /sdd:build, /sdd:check, /sdd:explain, /sdd:condense, /sdd:reorganize) over one SPEC.md. Telegraph encoding (~30% token cut vs Claude prose re-expansion).
Bug → spec protocol: trace cause, decide if a new §V invariant catches the recurrence class, append §B. Not user-invoked — engaged from /sdd:spec BACKPROP mode (bug → spec user route = `/sdd:spec <intent>`). Not for mechanical typo or one-off fix w/ no recurrence class — pure code fix only.
Plan-then-execute impl vs SPEC.md. Triggers when user asks to build, implement, or execute spec or specific §T task. Phrasings: "build §T.<n>", "build --next", "implement next task", "run the build", "does the implementation run?", "is §T.<n> done?".
Read-only drift detector. Diffs SPEC.md vs current code, reports violations grouped by severity. Writes nothing — suggests remedies via spec or build skills, never invokes them. Triggers when user asks to check drift, audit spec, or verify invariants. Phrasings: "check drift", "audit the spec", "check invariants", "spec vs code", "is the spec still accurate?", "did the code drift?".
SPEC.md condenser — token-budget sweep. Triggers when user invokes `/sdd:condense` or asks to condense spec or /sdd:check emits `## advisory` token-budget overflow line. Phrasings: "/sdd:condense", "condense SPEC.md", "SPEC too big", "shrink the spec", "token budget".
Propose-then-critique structural design loop → draft to `designs/<slug>.md`. Distinct from socratic (which sharpens vague intent). Use when user wants to design a structural change, weigh tradeoffs between named alternatives, propose an architecture, or shape a subsystem before implementation. Triggers: "/sdd:design", "design the X", "shape the X subsystem", "tradeoffs between A and B", "how should we structure", "propose an architecture for".
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.
Code consistency is the casualty of LLM agent velocity.
LLMs write code faster than any human can read it — and faster than the agent can stay coherent with itself. SDD keeps the spec small, dense, and durable. It's the part the agent re-reads every turn, so task ten is built against the same constraints as task one.
The mechanics:
§V.<n> / §T.<n> / §B.<n> are stable cites — code comments link to the invariant they uphold, tests reference the bug they guard, commits cite the task they close. SPEC.md survives /clear and team handoff.benchmarks/telegraph/). The savings come from terse grammar — dropped articles/filler, fragments, unpadded pipe tables — plus compact §-refs and a curated low-token symbol set (→ ≥ ≤ ! ? § |); Distinct from steno (the bundled human-facing shorthand).§B row, usually a new §V invariant. The drift report stays trustworthy because every prior failure tightened the spec.SPEC.md mutations, status flips, commits. Read-only audits (e.g. /sdd:check) may fan out to sub-agents. No orchestrator. Same spec + same task → same plan./sdd:check. You get a read-only drift report: which §V invariants the code violates, which §T tasks remain. No digging through old transcripts.The spec is the only artifact that earns its tokens. Everything else must save more tokens later, save the agent's context, or get cut.
SPEC.md is an LLM-facing artifact. You operate it through Claude — /sdd:spec writes, /sdd:build and /sdd:check read, /sdd:explain decodes a citation back to prose when you want to read along. The loop is human → /sdd:* → Claude → SPEC.md, not hand-editing in your editor.
That framing is load-bearing. Telegraphic fragments over full sentences, pipe tables over bulleted lists, dropped line citations — all optimized for the model that re-parses the spec every command, not the human skimming it. If you want to skim it as a human, /sdd:explain is the front door.
/plugin marketplace add kborovik/spec-driven-dev
/plugin install sdd@spec-driven-dev
Then in any repo:
/sdd:spec # creates SPEC.md if missing
┌────────────┐ ┌────────────┐ ┌────────────┐ ┌────────────┐
│/sdd:design │──►│ /sdd:spec │──►│ /sdd:build │──►│ /sdd:check │
│ propose │ │ mutator │ │ plan→exec │ │ read-only │
└────────────┘ └─────▲──────┘ └─────┬──────┘ └─────┬──────┘
│ │ │
│ ▼ on failure │ on drift
│ ┌────────────┐ │
│ │ backprop │ │
│ │ §B (+ §V) │ │
│ └─────┬──────┘ │
│ │ │
└────────────────┴────────────────┘
amend SPEC.md
SPEC.md. No docs/ tree, no JSON sidecars./sdd:spec. (/sdd:build may flip a . → x; nothing else writes.)/sdd:check (drift report) and /sdd:explain (decompression).Six fixed sections, fixed order. Each row is addressable as §<S>.<n>.
# SPEC
## §G GOAL
one line. what code must do.
## §C CONSTRAINTS
- non-negotiable boundary
- tech / language / library locked in
## §I INTERFACES
external surface — what the world sees.
- cmd: `foo bar` → stdout JSON
- api: POST /x → 200 {id}
- file: `config.yaml` schema …
- env: `FOO_KEY` required
## §V INVARIANTS
numbered. testable. each ! MUST hold.
V<n>: every req → auth check before handler
V<n>: token expiry ≤ current_time → reject
V<n>: DB write ! in transaction
## §T TASKS
id|status|task|cites
T<n>|.|scaffold repo|-
T<n>|.|impl §I.api POST /x|V<n>
T<n>|x|add §V.<n> middleware|V<n>,I.api
## §B BUGS
id|date|cause|fix
B<n>|2026-04-20|token `<` not `≤`|V<n>
B<n>|2026-04-21|race on write|V<n>
Status markers: . todo · x done.
Cell rules: literal | → \|. Empty cell = -. Backticks OK.
/sdd:design — propose-then-critiquenpx claudepluginhub kborovik/spec-driven-dev --plugin sddpilotkit — compressed spec-driven development. Three commands (/sdd:spec, /sdd:build, /sdd:check) over one SPEC.md. Caveman encoding. Bug-to-spec backprop. Adapted from JuliusBrussee/cavekit.
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.
Next.js development expertise with skills for App Router, Server Components, Route Handlers, Server Actions, and authentication patterns
Unity Development Toolkit - Expert agents for scripting/refactoring/optimization, script templates, and Agent Skills for Unity C# development
Design fluency for frontend development. 1 skill with 23 commands (/impeccable polish, /impeccable audit, /impeccable critique, etc.) and curated anti-pattern detection.
Behavioral guidelines to reduce common LLM coding mistakes, derived from Andrej Karpathy's observations on LLM coding pitfalls
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.