スキルを探す

Get told the moment a live Builder session is blocked (five refused submits in a row, a repeated findings set, twelve checks without acceptance, two environment previews in a row, two hours since a submit or rehearsal last returned) and fix the owner iteratively: read, fix on the owning PR, then keep the run or kill it and relaunch on the upgraded system, and rearm. Use while a paid run is live and the operator wants blocking caught and repaired, not only reported.

日本語の概要は準備中です。原文の説明を表示しています。

Launch, start, monitor, and collect independent Codex subagents (gpt-6-luna at high, xhigh or max; gpt-6.1-sol for small review batches) for bounded parallel work, from Codex or from Claude Code, including requests supplied as a Markdown file of session prompts. This skill owns Codex subagent transport even when another investigation or review skill defines the questions. Use whenever the user asks for Luna or Sol agents, a swarm, many sessions, a concurrency test, a particular reasoning effort, a Codex subagent from Claude Code, or later collection of reports. Distinguish launch-only requests from requests to wait, collect, or synthesise. Route high and xhigh through the direct launcher; for 16 or more sessions, also invoke the direct launcher instead of native spawn_agent.

日本語の概要は準備中です。原文の説明を表示しています。

Review every harness version the Builder produced for one domain. Join workspace commits to iteration evidence, saved execution versions, measured batteries, and claims. Read-only over controller output.

日本語の概要は準備中です。原文の説明を表示しています。

Check semantic interference when concurrent changes or stacked PRs meet. Use after a merge or rebase, at a stack checkpoint, or when comparing changed source with a live run. Measure overlap and exercise the affected combined path; stack-hop owns publication and failed-checkpoint isolation.

日本語の概要は準備中です。原文の説明を表示しています。

Use when the operator wants two or more model conditions (Opus 5, Fable 5.1, Sol) compared on one source head and one prompt, or wants a harness change that must serve more than one model. Covers the condition design that separates harness quality from solver capability, the recorded-evidence join, the reading of each bucket, and the rule under which a change is admitted for every model rather than one.

日本語の概要は準備中です。原文の説明を表示しています。

Audit and consolidate the Anabasis skill set. Compare current source, observed skill use and callers; move useful guidance to an existing owner and retire redundant entry points. Preserve main-versus-stack differences, scripts, independent review coverage and resolving links.

日本語の概要は準備中です。原文の説明を表示しています。

Test an uncertain Anabasis path at the smallest real layer that can decide it: deterministic joins, live authoring and admission, semantic host probes, or controlled model comparisons. Use full-run rehearsals only when the question needs them; not for ordinary run review or a path already proved by its owning test.

日本語の概要は準備中です。原文の説明を表示しています。

Invoke whenever writing, changing, reviewing, or sweeping tests in Anabasis. A write-time gate for every new or changed test, and a read-first audit that prunes or consolidates tests re-asserting source, duplicating a stronger owner, coupling to implementation, or keeping a test-only export alive. Knows the suite's helpers, fakes, walls, rule-4 invariance tests and runner, and hands removal claims that rest on overlap to test-impact-and-consolidation.

日本語の概要は準備中です。原文の説明を表示しています。

Use when designing, reviewing, or reporting Anabasis's three isolation layers, candidate validation, fingerprint, and immutable bundle identity. Covers executed refusal probes, combined access rules, candidate workspace access rules, typed public projection, controller-held authority, and honest disclosure of any boundary that is still contractual.

日本語の概要は準備中です。原文の説明を表示しています。

Select, explain and publish the strongest Anabasis runs from one calendar week. Use for the weekly best-run Automation, top-three/top-five comparisons, and questions about what made a run good, what limited it, or what should happen next. It filters by duration, reads every run through the controller's strict evidence and the outcome readers, reads published main syntheses, and challenges finalists with Luna xhigh.

日本語の概要は準備中です。原文の説明を表示しています。

Use when deciding which Anabasis stage owns a change, when carrying one bounded build/measure slice from the user's prompt to evidence, or when deciding what may enter a run: the exact one-line prompt, context files, public catalogues, research and solve-side public data. Maps the product loop, the input contract, the authoring sessions, the adoption gates, the four owners, and validate-or-reopen behaviour. Not a long-running campaign controller.

日本語の概要は準備中です。原文の説明を表示しています。

Use after an Anabasis run, comparison, or system change and before claiming improvement. Attributes movement to one owner, reports identities and censored denominators, separates deterministic proof from model judgement, keeps measurement validity apart from observed exploitation, and states what remains unrun or provisional.

日本語の概要は準備中です。原文の説明を表示しています。

Use for 1-7 independent read-only sessions investigating one bounded Anabasis failure, diff, or design question before a fix. Defines non-leading prompts, evidence packets, symmetric fault hypotheses, self-falsification, source adjudication, and concise synthesis. Do not use for whole-run coverage at any session count; use whole-run-investigation instead.

日本語の概要は準備中です。原文の説明を表示しています。

Shared vocabulary for designing deep modules. Use when the user wants to design or improve a module's interface, find deepening opportunities, decide where a seam goes, make code more testable or AI-navigable, or when another skill needs the deep-module vocabulary.

日本語の概要は準備中です。原文の説明を表示しています。

The commit-R protocol — test-driven rewrite of one compartment. Commit, remove the compartment's tests ("R:" commit), rewrite them from a stated hypothesis so production fails on purpose, remove the production files ("R:" commit), then rewrite production and patch or delete the outdated adjacent functions. Use when the operator says "commit R", "commit-R protocol", "remove and rewrite the tests", or asks for a TDD pass that looks for a faster, simpler path to the same outcome.

日本語の概要は準備中です。原文の説明を表示しています。

Design, change or audit one Anabasis contract: adoption gates, verifier discrimination, artifact representation and the tools over it (including whether a tool earns its place from recorded use), model-visible text, or coding discipline including a small rule-changing fix. Load only the relevant area; source and AGENTS.md own current mechanisms and authority.

日本語の概要は準備中です。原文の説明を表示しています。

Find the places in Anabasis where TypeSafe's Jev, a cheap calibrated System One classifier (Noul, Choice, Score over a text state), could replace or sit beside an LLM call, an embedding classifier, a prose regex, a lexical ranking or a blind-judge labelling loop, and prove each one in shadow against recorded labels before anything moves. Sweeps the tree for the five shapes, puts every candidate through the authority filter (what owns the decision, can it touch a pass, a score, a denominator or a gate) and the jev-1.13 fit filter, then writes a surface card with a runnable Bun experiment. Use when asked where Jev, TypeSafe or a System One model fits, what it could replace, or to run the autoresearch feature-discovery loop over this repository's records.

日本語の概要は準備中です。原文の説明を表示しています。

Launch authorised Anabasis runs or model pairs through one Bun TypeScript command, or stop an identified run under existing user authority. Owns preparation, exact condition checks, startup and stop mechanics; whole-run-investigation's outcome reference assesses whether stopping is justified.

日本語の概要は準備中です。原文の説明を表示しています。

Use when adding, changing or reviewing Anabasis's Main Judge: the battery review, its recorded disagreements and the advice they feed. Keeps the Judge outside verification, evidence-bound, separately reported, and unable to alter pass, acceptance, claimability or adoption.

日本語の概要は準備中です。原文の説明を表示しています。

Critically supervise another Codex or agent session without taking it over, or audit a long, compacted or delegated session for commitments it forgot, narrowed or left unverified. Maintain executable closure, material-claim, and intervention ledgers; challenge unsupported claims before decision boundaries; correct proved errors; and stop only after accepted work and owned processes close. Use for long-running tasks, campaign watches, delegated reviews, and likely-mistake checks.

日本語の概要は準備中です。原文の説明を表示しています。

Hand a question to an outside Oracle model (GPT-6 Pro or whichever the operator names) as one task document plus one zip. The Oracle reads GitHub and the harness-builder-v4 history; the zip carries what it cannot fetch: run traces via zip-run, patches for unpublished commits, notes, ledgers, predictions and plans. Use for "oracle prompt", "prompt for GPT-6 Pro", "gpt-6-pro appender", "zip the traces for the oracle", or the retired gpt-5-pro-appender.

日本語の概要は準備中です。原文の説明を表示しています。

Select, install, and name the authoritative compiler, simulator, solver, or other external verifier for an Anabasis harness. Use when deciding which practitioner tool to trust, writing the truth check that runs it, or diagnosing verifierUnavailable, timeout, crash or sandbox results. Also owns the install probe, the executable board for what the Builder can install and run behind the real cells and wall. Use harness-contract (discrimination proof) for semantic accept/reject coverage.

日本語の概要は準備中です。原文の説明を表示しています。

Generate and tune an AST-derived census of model-visible text in TypeScript or JavaScript. Use to map prompts, hooks, tool descriptions and feedback with their selecting conditions, verify delivery boundaries, and investigate prompt simplification. The census maps source candidates; recorded requests establish actual delivery.

日本語の概要は準備中です。原文の説明を表示しています。

Measure whether a change made the stack simpler, on numbers rather than impressions — line counts, branching, coupling, interface width, rule count and model-visible prose — and compare revisions or a whole PR stack on all of them at once. Use before claiming a pass reduced complexity, and when deciding which of several PRs added the machinery.

日本語の概要は準備中です。原文の説明を表示しています。