本文へ移動
cccskills
無料GitHub で公開

organize-processor-docs

Establish, author, restructure, or audit human-first processor Architecture, Design, Research, Review, and Verification documentation. Use for progressive documentation scaffolding, authority maps, reading paths, document-type content contracts, length-budgeted splitting, or maintainability reviews. Do not use as a substitute for cycle-accurate microarchitecture analysis or RTL implementation.

インストール方法を見る

含まれるファイル(13)

  • SKILL.md8.0 KB
  • agents/openai.yaml302 B
  • references/adr.md1.0 KB
  • references/architecture.md3.9 KB
  • references/bootstrap.md5.6 KB
  • references/design.md6.2 KB
  • references/maintenance.md5.7 KB
  • references/protocol-lifecycle.md3.9 KB
  • references/research-review.md2.2 KB
  • references/verification.md2.0 KB
  • scripts/check_docs.py31.6 KB
  • scripts/test_check_docs.py17.8 KB
  • scripts/test_skill_package.py5.7 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Organize Processor Docs

Build a directly editable Markdown document network that a processor designer can read without Harness state, a generated renderer, or a second processor model.

Authority

  1. Read the project AGENTS.md first.
  2. Treat explicit user decisions and the current Git authorities as authoritative. Treat a later user statement as change intent unless the user clearly supersedes an existing fact.
  3. For a new project without an established mapping, keep the current document network under one project-root doc/, matching bootstrap-processor-project. Use one current doc/Architecture/, one current doc/Design/, and one current doc/Verification/ when those domains contain real material. Use doc/Research/ only when project-maintained research exists. An existing project's approved mapping takes precedence. Git keeps history. Research, Review, and Finding remain evidence rather than processor authority.
  4. Give each normative fact one owning document. Summaries link to the owner and add no new normative detail.
  5. Keep Research, reference implementations, current RTL, current proposed documents, and new recommendations distinct.
  6. Do not introduce a document manifest, processor schema, renderer-owned truth, backup tree, or document workflow state.
  7. Do not create empty directories, empty topic files, or speculative placeholders.
  8. Explain every Chisel-facing interface in Scala declaration -> semantics order. Show the minimal Scala structure first, then explain fields in declaration order.
  9. Outside interface declarations, use Scala only when prose, a table, or a diagram cannot express the required hardware structure precisely. Keep only the minimal decisive fragment.
  10. Keep a same-stem editable source beside every explanatory raster diagram and update both in the same candidate change. Evidence captures such as waveforms and tool screenshots do not require an editable diagram source; bind them to the input commit, run or method, and evidence location.
  11. Use the physical module view in doc/Design/ as the main directory axis. Align this view as closely as possible with stable Chisel or RTL instance hierarchy and responsibility boundaries. Keep Protocols, Lifecycles, ADRs, and Verification as orthogonal views linked to that axis.

When detailed processor semantics are being designed or reviewed, also use design-chisel-processor. This skill owns information architecture and writing constraints; design-chisel-processor owns cycle-accurate correctness.

All doc/ paths in this skill and its references describe the default layout. For an established custom layout, resolve them through the project's mapping, including domain entries and the overall reading entry. Preserve that mapping, its authority files, and its local constraints. Do not silently rewrite AGENTS.md or create a parallel doc/ tree. A migration requires authorization and coordinated updates to the mapping, documents, and links.

Select a mode

Bootstrap

Use when a project lacks a clear document framework or when existing material must be organized. Read references/bootstrap.md.

Author

Use when creating or revising a document within an approved framework.

  1. For Architecture goals and processor properties, read references/architecture.md.
  2. For Design entry, overview, topology, subsystem, or module documents, read references/design.md.
  3. For cross-module Protocol or Lifecycle documents, read references/protocol-lifecycle.md.
  4. For ADR documents, read references/adr.md.
  5. For Verification documents, read references/verification.md.
  6. For Research, Review, Finding, or Diagnosis documents, read references/research-review.md.

Read only the references required for the current document types.

Maintain

Use when auditing, splitting, merging, relocating, or pruning existing documents. Read references/maintenance.md plus the references for every affected document type.

Common workflow

  1. Record the exact input commit or working-tree baseline and read the relevant authorities and evidence.
  2. Identify the reader question, owning fact, and shortest valid reading path.
  3. Preserve user terminology and topology unless the user approves a change.
  4. Propose authority or topology changes with affected paths before editing.
  5. Materialize approved work as readable Markdown and update entry maps and backlinks in the same candidate.
  6. Run deterministic checks and semantic review.
  7. Report changed paths, responsibility or reading-path changes, validation, and unresolved user decisions.

Provisional length budgets

Count effectiveChars as non-whitespace Unicode characters and nonBlankLines as lines containing non-whitespace content. Count prose, tables, code blocks, links, and embedded examples. Store raw logs, full source listings, and bulk research elsewhere.

Document kindTarget effectiveCharsTarget nonBlankLinesProvisional hard effectiveCharsProvisional hard nonBlankLines
Entry README.md2500604000100
Architecture topic, Design principles, Design overview600014010000200
Module, Protocol, Lifecycle800018012000250
ADR2500604000100
Research or Review600015010000220
Finding2500604000100
Verification specification600015010000220

These budgets are a writing strategy pending validation against real processor projects. Exceeding a target produces a warning and requires a concision review. Provisional hard-limit enforcement is configurable; use blocking enforcement only when the project or evaluation protocol explicitly selects it. Reduce duplication and incidental detail before splitting. Split only at a stable responsibility, ownership, protocol, lifecycle, or independent reading boundary. Never create Part1, Part2, or size-only fragments. Preserve necessary field tables, pipeline tables, state machines, and assertions when a justified document remains above a provisional threshold.

Human-first acceptance

  1. A human can understand every document without reading machine state or generated schemas.
  2. Every document opens with its responsibility, scope, and owned facts in natural language.
  3. doc/README.md links directly to every present domain entry.
  4. Any Architecture property is reachable from doc/Architecture/README.md within two document links.
  5. Any module, Protocol, or Lifecycle is reachable from doc/Design/README.md within two document links.
  6. The Design module view accounts for every stable implemented module responsibility and records justified differences from the Chisel or RTL instance hierarchy.
  7. An implemented Design document links its relevant Source and Test locations without copying source bodies.
  8. Verification material links to the Architecture property or Design invariant it checks.
  9. Every explanatory raster diagram retains a same-stem editable source. Evidence captures identify their input commit, run or method, and evidence location.
  10. Removing .assistant/ leaves the formal documentation complete and readable.
  11. Direct user edits remain first-class input and are not overwritten from another representation.

Use references/maintenance.md for deterministic commands, semantic audit, and change-report fields.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Initialize or safely upgrade a processor project's root AGENTS.md from a maintained baseline. Use when starting a processor project, adding project-level Agent collaboration rules, or comparing an existing AGENTS.md with the package baseline. This skill only handles AGENTS.md; it does not scaffold documentation, inspect or configure environments, install tools, or modify processor source.

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。

sixblade325/WaterHand-Processor-Development-Skills152026年10月3日 更新

Develop, challenge, and document Chisel processor microarchitecture designs before implementation. Use when discussing or writing design documents for processor pipelines, queues, issue logic, rename, ROB, LSU, caches, MSHRs, forwarding, wakeup, replay, flush, privilege, exceptions, or other cycle-accurate hardware mechanisms; when converting design conversations into stable Markdown specifications; or when reviewing a proposed Chisel CPU design for correctness, timing, area, verification cost, and cross-document consistency.

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。

sixblade325/WaterHand-Processor-Development-Skills152026年10月3日 更新

Document-driven workflow for implementing, reviewing, and verifying Chisel processor and memory-subsystem RTL. Use when Codex must work from maintained Architecture and Design, trace a complete integration surface, reason in synthesized-hardware terms, keep source-adjacent _codex.md summaries current, avoid redundant or overprotective logic, or run focused functional 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。

sixblade325/WaterHand-Processor-Development-Skills152026年10月3日 更新

Diagnose and optimize timing-critical Chisel RTL for FPGA implementation while preserving cycle semantics. Use for Vivado timing bottlenecks, long ready or admission paths, queue and issue selection, free-list bank mapping, priority encoders, one-hot arbitration, wide muxes, late-arriving forwarding or override data, high-fanout controls, cross-module predicates, register-boundary changes, emitted-Verilog inspection, or routed-DCP A/B analysis. Also use when a source-level simplification needs proof that it changes timing without changing architectural state.

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。

sixblade325/WaterHand-Processor-Development-Skills152026年10月3日 更新

Task-sized forensic analysis of Vivado synthesis and routed timing evidence for processor RTL. Use when Codex must trace a named setup or hold path, audit whole-design timing populations, compare implementation runs, map primitive, LUT, CARRY, BRAM, DSP, MUXF, register, and routed-net stages back to Chisel or generated RTL, distinguish logic from routing pressure, identify missing DCP queries, or write an evidence-backed timing report. This skill is read-only by default and stops at ranked modification directions. Use optimize-chisel-fpga-timing for RTL edits and routed A/B implementation closure.

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。

sixblade325/WaterHand-Processor-Development-Skills152026年10月3日 更新

sixblade325 のスキルをすべて見る

このスキルの問題を報告する