「step」の検索結果

Build AI agents with Hugging Face's SmolAgents framework. Use when creating code-executing agents, tool-calling agents, multi-agent systems, agentic RAG, text-to-SQL pipelines, web browsing agents, or any multi-step AI workflows. Covers CodeAgent, ToolCallingAgent, custom tools, MCP integration, memory management, secure code execution (E2B, Docker, Blaxel), and model configuration (HF Inference, LiteLLM, Transformers, Ollama).

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 orchestrating implementer/judge subagents — form gate + nine modes (do-and-judge ±two-stage, steps/parallel/worktrees, competitively, debate, live-app-judge, adversarial-council).

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 reviewing a README for accuracy, usability, and alignment with the actual repository. Detects invented content, broken setup steps, and structural issues.

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 multi-step / ambiguous / end-to-end work — refactor a whole module, drive a vague ticket to a verified result, plan+build+verify a migration; coordinates the reasoning chain across skills.

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 debugging PHP with Xdebug — breakpoints, step-through, dual-container setup, IDE configuration, header-based routing — even when the user just says 'why does this blow up on request X'.

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 feature is implementation-complete and the next step is 'ship it' — verifies, cleans up, routes to merge/PR/park/discard; never destroys work without explicit confirmation.

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 creating or editing a slash command in src/agent-src/commands/ — frontmatter, numbered steps, safety gates — even when the user just says 'add a /command for X'.

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 what earns a page — the page, action and info classes stated provider-neutrally, and the rule that a page without an owner, a runbook and a first diagnostic step is malformed.

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。

Controle saída de LLM com regex e gramáticas, garanta geração válida de JSON/XML/código, aplique formatos estruturados e construa fluxos de trabalho multi-etapas com Guidance - framework de geração restrita da Microsoft Research

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。

[AI & Tools] Use when you need to apply structured, reflective problem-solving for complex tasks requiring multi-step analysis, revision capability, and hypothesis verification.

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。

[Process] Use when the developer wants something explained or taught, or wants to know HOW TO REVIEW a change — by default the current working tasks + changes in context, or whatever the prompt names (a plan, a subsystem, a decision, a concept, a bug). AI derives WHAT to explain from the prompt and ALWAYS delivers a teacher-and-coach review guide, high level first then detail: what was done, a VISUAL MAP (system flowchart, domain ERD, sequence diagrams), the user stories and business rules with their REAL test-case IDs, a REVIEW PATH naming which files to open first and in what order with an exit criterion per stage, the technical concepts needed to follow it, how it works, why THIS solution, every alternative option with pros/cons, the trade-offs accepted, the blast radius, HOW TO TEST AND DEMO the change, the developer's decision levers, and written challenge prompts that provoke the reader to pressure-test the work. Works at ANY scope — from one line of code to the whole project: it SIZES the target into a scope tier, DECOMPOSES a large one into understanding groups, BREAKS the work into tasks before the first deep read, and ACCUMULATES the report group by group on disk so a huge explanation stays complete instead of thinning out. Writes the full report to a single combined markdown file (plans/reports/understand-*.md) that opens with a detailed summary section — or delivers it in full in chat when Step 3 finds no git-ignored directory for it — and always summarizes it in chat. Regardless of coding level. Never interrogates, never quizzes, never blocks.

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。

[Implementation] Use when you need to start coding & testing an existing plan. Flags: --approval=off (auto/trust mode, no approval gate), --tests=off (skip the test step), --parallel={auto|on|off} (default off — sequential; --parallel/=on opts in to parallel sub-agent waves; =auto fans out only when the plan declares PAR/SEQ tags and write sets).

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。

[Utilities] Use when the user invokes, lists, saves, updates, or deletes a PROJECT-SPECIFIC saved prompt — a named, reusable procedure stored with the project rather than in the portable framework. `/custom-prompt list` prints every defined prompt with its description; `/custom-prompt <free-text request>` matches the request to the closest saved prompt and executes it only after a confirmation gate; `/custom-prompt save|update|delete ...` maintains the registry — save never stores raw wording, it drafts the best version of the prompt and confirms the name, description, inferred goal, and steps with the user first. Triggers: custom prompt, custom prompts, list custom prompts, save this prompt, save this prompt task, update this prompt, my saved prompt, project prompt, playbook, recipe, runbook.

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。

Run a Mixture of Experts (MoE) audit on any topic, plan, codebase, evidence archive, or process. Dynamically spawns specialized subagents with different critical lenses to cross-reference findings and detect flaws, overlap, retention issues, or drift. Use when designing architectures, analyzing complex code, verifying multi-step plans, classifying evidence artifacts, or looking for duplicated intent in a repo.

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 between coding-agent runs (Codex, Claude Code, Cursor, Aider, etc.) to stop the project from drifting, overbuilding, or accumulating dead weight. Forces a read-only audit that classifies every active item as KEEP / DELAY / REMOVE and ends with a single, unambiguous next action. Trigger on "audit before continuing", "what should I do next", "this project got too big", "between runs", "trim scope", "stop adding stuff", or any signal that the project feels noisy and the next step is unclear. Do NOT load for trivial one-line fixes, when the next action is already obvious, or when you're mid-run and just need to keep going.

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 you have a spec or requirements for a multi-step task, before touching code

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 writing or changing a model-facing tool definition (name, description, parameter schema, or its system-prompt section), and when designing skills, context loading, or multi-step workflows, to make information discoverable and use context efficiently.

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 reusable multi-step agent workflow blueprints. Use for trigger/action orchestration, deterministic workflow definitions, and automation handoff artifacts.

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。

Start a project on the Visual Stack — a multi-step setup form captures what's being built, recommends a stack, and either instantiates the vstack-template-base template, sets up a specs-and-design-only workspace, or records an existing codebase without touching it. Use when the user wants to start a new project, scaffold a repo, set up a codebase from the Cavalry template, pick a stack or add-ons, run Day-1 setup, or bring an existing app onto the visual stack.

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。

Produces a deep, honest peer-review of how someone uses Claude Code by statistically analyzing their local session data (~/.claude/usage-data/ and ~/.claude/projects/). Trigger whenever the user asks to review, analyze, audit, or critique their own Claude Code usage, workflow, habits, or skill level — including phrases like "analyze my cc usage", "review my cc sessions", "peer review my cc workflow", "deeper than /insights", or any ask for an honest audit of their AI workflow. Also covers portfolio/hiring-manager framings (e.g. "portfolio for Anthropic/OpenAI/xAI"), but ALWAYS ask the user in Step 0 which version to build (self / hr / both) before running — never silently produce an HR report, because that version goes to outsiders and needs explicit privacy setup. The HTML report is laid out ledger-first for SELF (opening band, output/team/leak/trend ledgers, then peer review, 9-dim scoring, try-this-week, case study, claim-indexed evidence). The HR variant is a recruiter version v1: identity card, earned badges (threshold-based, earned-only), allowlist-filtered output ledger, one case study, and a methodology & scope disclosure — no scores, no peer-review memo, no charts.

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。

Durable categorized record-keeping for any predicate-enabled walk. Trigger when: - Classifying a step's output for recording (decision, finding, debt, feedback). - Promoting a scratch artifact to the ledger, to repo/docs, or to a category home (.ledger/tech-debt/, .ledger/process-feedback/). - Auditing whether a draft is structurally ready for its destination. - Running a zero-context sensibility read before committing to promotion. - Recording a known limitation as a tech-debt entry (not prose in the log). - Recording a friction, miss, improvement, or process amendment. - Prompt contains: /record, record-keeping, promote, ledger note, zettelkasten, scratch-to-ledger, zero-context read, over-strip, classification trigger, tech-debt, process-feedback, amendment.

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。

SOP for the micro-execution C.O.R.E. phase. Trigger when: - Implementing plan steps, managing code changes at commit boundaries, and verifying incremental progress. - Navigating states: Absorb, Clarify, Plan, Execute. - Prompt contains: /core, core workflow, absorb, clarify, execution invariants, commit boundary, verify assertion.

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。

SOP for normative specification and behavioral contracts. Trigger when: - Defining invariants, permitted transitions, and forbidden states. - Navigating steps: Identify, Formalize, Verify, Record, Connect. - Prompt contains: /spec, spec workflow, normative constraint, invariant, BCP 14.

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。

A pessimistic go/no-go the moment you pick up a GitHub issue — read it skeptically and hunt for reasons NOT to do it (already done? obsolete? duplicate? premise wrong? cheaper way? net-negative?), returning proceed / reshape / drop with one-line evidence. The pickup-time half of the dedup gate (the create-time half is /issue-dedup). Use as step 1 of the Task Execution Workflow, or run /issue-sanity-check

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。