jawsublime-byte

Test how a trusted program ages under realistic accumulated use at 3 months, 6 months, 1 year, 5 years, and beyond. Use only when explicitly invoked near build completion to expose storage growth, stale state, latency creep, indexing cost, log or cache bloat,…

Inspect an in-progress build for demonstrably required pieces that were never named but are necessary for the user's approved final result. Use only when explicitly invoked to find empty slots before integration or testing makes the build fail, while…

Govern repeated builder drift with evidence-based yellow cards, a three-packet red-card benching period, reinstatement on probation, and expulsion from the current build after repeated post-return violations. Use only when explicitly invoked for multi-builder…

Run an evidence-grounded adversarial brainstorming session when the user's unconventional build idea conflicts with the model's learned defaults or standard practice. Use only when explicitly invoked to keep both positions intact, challenge assumptions…

Execute an identified build order exactly as written or explicitly route a builder problem to one bundled playground skill. Use only when the user invokes Simon Says and supplies an authoritative order or asks Simon Says to route a task. Do not activate for…

Investigate resource growth, retained state, data or information loss, dropped records, missing requirements, lossy transformations, unbounded accumulation, or other accounting failures across a bounded system or workflow. Use when inputs, resources, records,…

Evaluate competing, sequential, or evolving paths toward a defined goal and classify which paths advance progress, cause regression, contain useful partial progress, or remain unresolved. Use for workflows, architectures, implementations, product evolution,…

Use when a Rust implementation, review, repair, or Cargo decision requires authoritative syntax, language, edition, ownership, safety, workspace, crate, or build-system guidance from the approved local manuals, including when a Simon Says packet explicitly…

Enforce a strict read-only verifier boundary when an independent reviewer must inspect a completed or candidate build without repairing it, inheriting the Builder's reasoning, inspecting future work, or communicating fixes directly to the Builder. Use only…

Route a user-described builder problem to exactly one bundled nursery-rhyme or story-based skill. Use only when the user explicitly invokes Mother Goose and asks for routing. Do not execute every story or silently activate during ordinary work.

Challenge groupthink, authority bias, sycophancy, and yes-man consensus when agents or reviewers may be agreeing because a position is already approved or socially reinforced. Use only when explicitly invoked to test whether dissenting evidence or an…

Calibrate repeated verifier, security, or monitoring warnings so unsupported alarms do not drown out real evidence. Use only when explicitly invoked with a set of current or supplied alerts to assess.

Stop repeating the same operation under unchanged conditions after the same failure signature is reproduced. Use only when explicitly invoked for blind retries; use the user's retry budget or default to three identical reproduced failures.

Protect proven valuable behavior before optimizing, refactoring, simplifying, or replacing the component that produces it. Use only when explicitly invoked around a known-good capability whose value must survive the change.

Leave concise recoverable breadcrumbs through a long or multi-agent build so work can resume from evidence instead of memory. Use only when explicitly invoked to establish or inspect an operational checkpoint trail.

Attribute significant work, evidence, and verification to the actor or source that actually performed or supplied it. Use only when explicitly invoked for multi-agent work, handoffs, assembled artifacts, or ambiguous provenance.