skills

Audit the structural and logical integrity of a mathematics paper — quantifier scope, domain-of-definition guards, proof/computation honesty, theorem hierarchy, contribution-list discipline (mapping every claimed contribution to its formal result, prior baseline, concrete gain, and independent significance), stale claims, notation accuracy at introduction, standing assumptions, and undefined relation symbols or coined terms in summary prose — without touching prose style or rhetorical prose inflation (structural inflation of contribution lists and result hierarchy is in scope).

Audit a LaTeX mathematics document for notation drift — symbols used under different names across sections, orphaned macros, undeclared aliases (including cross-language pairs where a term defined in one language is later used in another), symbols or relation signs used with no definition site anywhere, symbol reuse across scopes, and first-use-after-gap without back-reference — without evaluating mathematical correctness or prose style.

Review or rewrite mathematical prose so an edit, paraphrase, terminology migration, or explanatory description preserves the exact mathematical meaning fixed by its semantic source — referent identity, operation versus result, domain/codomain and representation level, quantifier and equality mode, role and provenance, notation role, and classified-before-replacement terminology migration with concept-cluster synchronization — without auditing claim-versus-proof strength (use math-claim-integrity), notation bookkeeping (use math-notation-consistency), Japanese style or translation events (use wabun-math-style), or audience prerequisites and ordering (use exposition-flow).

Detect and correct Japanese-language anti-patterns in mathematical writing — epistemic hedges weakening proved claims, incorrect verb tense, passive/active confusion in proofs, ambiguous particles, unjustified 明らか, decorative connectives, redundant meta-discourse, double negation, stacked の, generic/general conflation, borrowed non-mathematical vocabulary, vague category names for explicit formulas, full-width semicolons or untranslated English, canonical-terminology drift, and translation-level concept conflation — without evaluating mathematical correctness or structural hierarchy.

Turn repository-grounded source material into a self-contained, audience-admissible document or presentation by modelling the intended audience, extracting information units with explicit discourse roles, and enforcing dependency-ordered exposition in which every load-bearing statement follows the prerequisites needed to interpret it — without factual verification or citation checking (use sot-integrity), prose polishing (use deslop-prose), mathematical-correctness review (use the mathematical-writing skills), or typography and slide layout.

Workspace-level Git cleanup — discover repositories under the current directory and prune merged branches, merged linked worktrees, stale remote-tracking refs, and stale worktree metadata.

Audit a source-of-truth artifact for authority, evidentiary grounding, trust scope, and conflict with repository reality before implementation or orchestration relies on it.

Decide whether a development task should be done now, handled with `lite-spec`, or sent to `metaplan`.

Review a candidate external skill against the repo policy, record approved entries in a project-local catalog, and recommend APM install commands when approved.

Decide whether to execute a task directly or split it into a small number of tightly bounded subtasks, and produce minimal subtask contracts only when orchestration is justified.

Generate a compact handoff prompt for the next agent or stage. Output the prompt only.

Review and tighten specs, plans, and task breakdowns before implementation so an autonomous coding agent can execute without ambiguity or avoidable re-research.

Clean final user-visible artifacts produced from accepted decisions, reviewed direction, or issue/PR discussion by removing non-essential historical context, discussion provenance, superseded alternatives, and process framing while preserving current decisions, contracts, constraints, operational rationale, and required compatibility notes.

Clean academic or professional prose that carries AI-generated residue by removing hype, decorative structure, vague metadiscourse, pseudo-technical phrasing, and claim-evidence mismatch while preserving technical meaning, qualifications, and authorial intent.

Assess whether a mathematical or applied-mathematical research direction is locally significant — separating correctness, novelty, and scholarly value — before prose polishing or source-of-truth treatment.

Use when writing, refactoring, or maintaining Lean / Mathlib proofs — especially across a large formalization — so soundness survives both the mechanical checks and a close reading of every statement, and progress is measured by discharged obligations rather than the appearance of motion.

Use when running, reviewing, or preparing computational experiments so inputs, outputs, manifests, failure states, and generated artifacts remain explicit and reproducible.

Issue-first, PR-gated delivery — no direct main pushes, independent review required, deterministic merge gates. Use whenever a task implements a GitHub issue and ships through a PR (assigned by an orchestrator, by project rules, or self-invoked by the implementing agent).

Seed, lint, or update a compact canonical AGENTS.md with deterministic guardrails. Use this when AGENTS.md is missing, stale, structurally degraded, or when durable repo guidance changed.

Rate-limit-aware GitHub Projects v2 board maintenance — REST-first candidate discovery, single GraphQL snapshot, mechanical status-update planning, guarded apply. Use whenever a request asks to organise a Project board for issues/PRs updated within the last N hours.

Use when a task explicitly requires random selection, probabilistic sampling, stochastic tie-breaking, or diverse creative generation. Prefer the bundled PRNG script when tool execution is available. Use prompt-only String Seed fallback only for low-stakes or creative-diversity cases. Do not use for security-sensitive, scientific, benchmark, lottery, or correctness-critical randomness.

Operational guidance for using Claude Code as a delegated worker behind a coordinator agent. Covers delegation posture, stable session identity, permission strategy, worktree/tmux patterns, dynamic agents, effort selection, bare-mode auth, MCP trust, and remote-control caveats.

Produce a compact execution spec for bounded, non-trivial implementation work that needs a short execution brief but does not require `metaplan`.

Use when running textlint from the command line in a repository that has, or is about to add, textlint configuration. Covers detection, repo-local invocation, fix mode, diagnostic flags, CI-oriented formats, MCP mode, and how to distinguish lint failures from tooling failures. Do not use to define a writing policy or to claim correctness beyond the configured rules.

Use for budgeted, structure-aware previews of large or hard-to-inspect JSON/YAML, logs, code, or notebooks before deciding what to inspect next with `hson`.

Use for non-trivial Bash/sh creation, edits, or review where correctness depends on execution context, permissions, filesystem mutation, argument parsing, idempotency, CI/container/devcontainer behaviour, or repeated shell-script review failures.

Use this skill when working with Dev Containers, devcontainer.json, or Dev Container CLI for setup, review, validation, or troubleshooting.

Review an existing marimo notebook for reactive design quality, identify anti-patterns against marimo's dependency model, and propose a concrete refactor plan without rewriting the notebook unless asked.