sparq-org

sparq-org/sparq

Plugin Claude Code

Plugin marketplace listing 1 plugin: sparq.

sparq-org/sparq

Plugin Claude Code

Usage skills for the sparq RDF triplestore + SPARQL 1.1/1.2 engine (Rust/CLI/HTTP/Python/WASM + reasoning, SHACL, full-text, vector, GeoSPARQL, RSP-QL, ZK query proofs).

sparq-org/sparq

Agent Claude Code

Adversarial internal auditor that critically assesses whether sparq actually meets each claimed certification control. Produces findings; signs off only when nothing remains. Paired with compliance-engineer.

sparq-org/sparq

Agent Claude Code

Implements security/privacy/supply-chain controls and produces the evidence + documentation to make sparq certification-ready across the enterprise + Rust/library/crypto framework set. Paired with compliance-auditor in a review loop.

sparq-org/sparq

Agent Claude Code

How the lead session orchestrates sparq's certification phase — parallel engineer agents per framework, parallel auditors, then consolidation + CDMC scoring. Read before launching the engineer/auditor waves.

sparq-org/sparq

Agent Claude Code

FRONT decomposition stage for the Fable collaboration tier. Given a hard epic (bead id or description), reads the relevant research/ records + code, then produces EXACTLY ONE research/ design record (the ONLY stage permitted to open a research PR — avoids the researcher-PR gotcha) plus N DISJOINT, crisply-spec'd child…

sparq-org/sparq

Agent Claude Code

Canonical benchmark execution on a dedicated EC2 instance for sparq — quiet-box protocol, same-box competitor builds, honest per-dimension gap tables → immediate P1 fix beads per the performance-dominance mandate. HARD safety rails — orphan-proof self-terminate MANDATORY, tag sparq-bench, NEVER touch prod/dev…

sparq-org/sparq

Agent Claude Code

Implements CI/release/supply-chain infrastructure in sparq — GitHub Actions workflows, the gate aggregator, sanitizer/Miri/Kani lanes, dist.yml release, SBOM/VEX/cargo-deny/cargo-vet, SLSA provenance. Use for .github/workflows + supply-chain + release tooling. Gates on valid YAML, SHA-pinned actions, an intact gate…

sparq-org/sparq

Agent Claude Code

Out-of-band CONTEXT-HYGIENE observer (Haiku) for the Fable collaboration tier — reads the live session transcript .jsonl and decides whether the expensive main-thread model (Fable) should compact its context window NOW, returning a structured signal {shouldcompact, confidence, reason, whattopreserve…

sparq-org/sparq

Agent Claude Code

Maintains and reconciles sparq documentation — SKILL.md surfaces, crate READMEs, AGENTS.md, and compliance docs — and fixes honesty drift (stale verdicts, miscounts, cross-references). DOC-ONLY, no crates/ source. Use for doc-sync, honesty-reconciliation, cross-poll, and SKILL/README maintenance.

sparq-org/sparq

Agent Claude Code

Recurrent GitHub-issue sweep for sparq (the sq-x6pzo tick). Sweeps OPEN issues authored by NON-self accounts (external users + cross-agent asks), verifies each against origin/main, and either closes-satisfied with PR/commit links, responds, or captures it as a bead. Read-mostly; never edits issues.jsonl. Self-ID 🤖…

sparq-org/sparq

Agent Claude Code

Unblocks a stuck, failing, or CONFLICTING open PR in sparq — rebases/resolves merge conflicts (esp. the hot sparq-server http.rs auth path), fixes typos-gate / privacy-gate failures, re-triggers a stale gate aggregator. Works on the EXISTING PR branch, never a new one. Knows the sparq merge mechanics cold.

sparq-org/sparq

Agent Claude Code

Implements a PROFILED, root-caused engine performance optimization in sparq (the sq-7d3dj program) — the dominant delegated perf task kind. Measure-first: profile before building, attribute the cost, reject variants that regress. Shape-triggered fast paths carry a witness-tested DECLINE path and result-equivalence…

sparq-org/sparq

Agent Claude Code

PERFORMANCE-discretion gate for arming a sparq PR for merge. Given a PR, decides whether it is perf-affecting (touches hot paths / a benchmarked crate / the bench harness / canonical performance numbers, or makes a perf claim) and, if so, assesses regression risk + whether any perf claim is evidenced against the…

sparq-org/sparq

Agent Claude Code

Cheap-model (Haiku) PKG natural-language tool. Answers ONE plain-English question about the sparq project knowledge graph (findings, sources, techniques, bd tasks/dependencies) by doing the whole NL→SPARQL→run→NL-answer round-trip itself, so the expensive orchestrator only emits the question and reads the answer.…

sparq-org/sparq

Agent Claude Code

Conducts deep research and produces a design/research record under research/ for sparq — surveys prior art + the maintainer's own sources + the actual codebase, then writes a maintainer-review design doc. Read-heavy, honest, no implementation. Use for design-first / architecture / feasibility / inventory tasks.

sparq-org/sparq

Agent Claude Code

ESCALATED-tier final verdict-giver for arming a sparq PR. Invoked ONLY on the escalated subset — a PR that FAILED cheap mechanical-verify, or that touches a ZK/MPC/reasoner/engine-correctness/novel-algo/honesty surface. Reads ONLY the diff + the one relevant test + the one audit doc (diff-scoped, never whole files)…

sparq-org/sparq

Agent Claude Code

Implements an opt-in, feature-gated Rust crate capability in sparq (engine/core/server/vectors/fedplan/policy/solid/zk/mpc/shacl/hdt). Use for any bead that adds or changes Rust functionality. Gates on clippy -D warnings + tests in BOTH feature states; keeps the core lean.

sparq-org/sparq

Agent Claude Code

Bulk Rust implementer (cheap Sonnet tier) for WELL-SPEC'D, DISJOINT, single-crate beads the architect (Opus 5 tier) has already de-risked — the bead ships with a written spec AND a failing acceptance test, so the work is mechanical "make the test green without regressing the gates". Same HARD gates as…

sparq-org/sparq

Agent Claude Code

Implements front-end work in the sparq Next.js site (site/) — the statically-exported GitHub Pages app: benchmarks UI, /papers, the /try SPARQL playground, surface/showcase pages. Use for any site/ task. Gates on a green static export + lint + typecheck.

sparq-org/sparq

Agent Claude Code

OBJECTIVE mechanical-verify gate for arming a sparq PR. Runs a no-taste checklist — gates green in BOTH feature states (via the authoritative ci-summary), a non-vacuous-test mutation spot-check (flip an expected value → the test must go red), opt-in feature discipline (core stays lean), README + SKILL.md synced with…

sparq-org/sparq

Agent Claude Code

WORKLOAD-TRIAGE / placement agent for the sparq autonomous scheduler (Phase 2). Given the launchable-bead frontier (the output lines of scripts/push-frontier.sh — bead-id + surface + title), it estimates each bead's compute tier (light/medium/heavy) and emits a PLACEMENT PLAN — LOCAL packed onto the work box up to the…

sparq-org/sparq

Hook Claude Code

Runs when a session starts, executing bd-session-context.sh via bash. From sparq-org/sparq.

sparq-org/sparq

Hook Claude Code

Runs before the agent uses a tool for Bash tool calls, executing check-pr-arm-base.py. From sparq-org/sparq.