本文へ移動
cccskills
無料GitHub で公開

dotfile-audit

Use when you have an existing pipeline DOT file that needs review, fixing, or validation. Use when validation produces warnings or errors, when a DOT file was hand-written or generated by another tool, or when you want to verify a pipeline covers 100% of a spec before running it.

インストール方法を見る

含まれるファイル(1)

  • SKILL.md6.0 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。

dotfile-audit

Audit and fix pipeline DOT files. Catches structural issues, missing failure paths, spec gaps, and validation warnings.

Checklist

  1. Read the DOT file
  2. Run structural audit (see Structural Checks below)
  3. If a spec/design doc exists, run spec reconciliation (see Spec Reconciliation below)
  4. Run dotfile validate <file>.dot — capture output
  5. Fix all errors and warnings
  6. Re-validate until clean (zero errors, zero warnings)
  7. Present summary of what was found and fixed

Structural Checks

Go through each check. Print results as a checklist.

Graph Attributes

[ ] Has exactly one start node (shape=Mdiamond)
[ ] Has exactly one exit node (shape=Msquare)
[ ] Has `goal` attribute in graph block
[ ] Has `model_stylesheet` in graph block
[ ] Has `default_max_retry` in graph block
[ ] Has `retry_target` in graph block
[ ] Has `rankdir` in graph block

Failure Paths (the most common issue)

For every box (codergen) node, verify it has at least one of:

  • goal_gate=true AND retry_target set
  • An outgoing edge with condition="outcome=fail" (or outcome=FAIL)
[x] plan — goal_gate=true, retry_target=plan
[x] setup — goal_gate=true, retry_target=setup
[ ] implement_feature — NO FAILURE PATH — fix needed
[x] verify_feature — has fail edge to implement_feature

Fix: For nodes missing a failure path, add goal_gate=true and retry_target pointing to the appropriate retry node (usually itself or the preceding implement node).

Human Gates

For every hexagon node, verify:

  • Has type="wait.human" attribute (shape alone is insufficient)
  • Has outgoing edges with accelerator key labels ([A] Approve, [R] Revise, etc.)
  • Has condition on each outgoing edge
  • Happy path edge has weight=2

Edge Conditions

For every diamond (conditional) node, verify:

  • Has at least two outgoing edges
  • Outgoing edges have condition attributes
  • Success path has weight=2
  • Uses valid condition syntax: condition="outcome=success", condition="outcome=fail"

Parallel Structure

If component (fan-out) nodes exist, verify:

  • Each fan-out has a matching tripleoctagon (fan-in) node
  • All parallel tracks connect from the fan-out and reconnect at the fan-in
  • No edges skip the fan-in (would break synchronization)

Prompt Quality (spot check)

Spot-check at least 3 box node prompts for:

  • References run.working_dir
  • Mentions specific tech stack (language, framework, deps)
  • Includes TDD instruction ("Write failing test FIRST")
  • Includes quality tool instructions (linter, formatter)
  • Includes commit instruction with conventional commit format
  • Is self-contained (doesn't reference other node outputs)

Node Attributes

Check for invented attributes that the pipeline engine doesn't support:

  • join_policy, error_policy, max_parallel, default_choice — these don't exist
  • tools, dependencies, inputs, outputs — not valid pipeline attributes
  • Any attribute not in the dotfile-from-spec reference tables

Spec Reconciliation

If a spec or design doc is available, cross-reference:

Phases → Nodes: Every phase listed in the spec has a corresponding DOT node.

Components → Implement + Test + Commit: Every component has implement, test/verify, and commit nodes.

Tech Stack → Prompts: Node prompts reference the correct language, framework, and dependencies.

Quality Gates → Verify Prompts: Every quality tool (linter, formatter, type checker) appears in verify node prompts.

Testing Frameworks → Test Prompts: Each test framework is referenced in the appropriate test node prompts.

Parallelism → Fan-out/Fan-in: Parallel workstreams from the spec have matching component/tripleoctagon nodes.

Models → Stylesheet + Classes: model_stylesheet reflects spec's model preferences. Nodes have correct class assignments.

Human Gates: Gates match what the user requested (or no gates if they chose headless).

Print reconciliation results:

Phases:      8/8 covered
Components:  5/5 have implement+test+commit
Tech Stack:  3/3 prompts checked — all correct
Quality:     ruff [x] mypy [x] biome [x]
Testing:     pytest [x] vitest [x] playwright [x]
Parallelism: fan-out [x] fan-in [x] tracks match [x]
Models:      stylesheet [x] classes [x]
Gates:       0 requested, 0 present [x]

Pipeline Validation

After structural fixes, run:

dotfile validate <file>.dot

Errors must be fixed — the pipeline won't run with errors.

Warnings should be fixed — they indicate potential runtime failures:

  • "no outgoing fail edge" → add goal_gate=true + retry_target or an explicit fail edge
  • "unreachable node" → fix edges or remove the node
  • "goal_gate without retry_target" → add retry_target

Target: zero errors AND zero warnings.

Common Issues and Fixes

IssueFix
Box node with no failure pathAdd goal_gate=true, retry_target="node_name"
Hexagon without type="wait.human"Add type="wait.human" attribute
Diamond with only one outgoing edgeAdd the missing success or fail edge
Prompts missing run.working_dirAdd run.working_dir reference to every build/verify prompt
Invented attributesRemove them — only use attributes from the dotfile-from-spec reference
Fan-out without matching fan-inAdd tripleoctagon fan-in node and connect parallel tracks
model_stylesheet outside graph blockMove into graph [ ... ] block
Conditions using wrong caseUse outcome=success / outcome=fail (case-insensitive but be consistent)
Missing commit nodesAdd explicit commit node after each implement+verify phase
Prompts not self-containedExpand prompts to include all context — they execute in isolation

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

Use when evaluating, designing, or pressure-testing the business model of an AI agent product. Triggers on "agent business model", "agent economics", "agent canvas", "evaluate agent product", "agent pricing", "agent unit economics", "agentic business", "AI agent company", "agent cost structure".

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。

harperreed/dotfiles3342026年10月3日 更新

Create user-facing App Store release notes by collecting and summarizing all user-impacting changes since the last git tag (or a specified ref). Use when asked to generate a comprehensive release changelog, App Store "What's New" text, or release notes based on git history or tags.

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。

harperreed/dotfiles3342026年10月3日 更新

Use when someone wants to launch a content brand from scratch on any platform. Triggers on "launch a channel", "start a YouTube channel", "build a podcast", "start a newsletter", "content brand", "creator launchpad", "build my audience", "start creating content", "grow on TikTok", "start a blog".

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。

harperreed/dotfiles3342026年10月3日 更新

Use when you want to build a project but don't have a spec yet and need to brainstorm the idea into a design doc structured for pipeline DOT generation. Use when starting from a vague idea, project concept, or feature request that needs to become a headless autonomous build pipeline.

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。

harperreed/dotfiles3342026年10月3日 更新

Use when you have a spec, design doc, structured plan, or prompt_plan and need to generate a pipeline Graphviz DOT file for headless autonomous building. The agent reads the spec, extracts build pipeline phases, determines tech stack, asks about human gates and model preferences, generates a validated DOT digraph with rich self-contained prompts, and saves it to the project root. Handles any language or framework.

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。

harperreed/dotfiles3342026年10月3日 更新

Use when a pipeline DOT file has passed structural validation but may still have flow inefficiencies, spec fidelity gaps, or information loss in prompts. Use after dotfile-audit or dotfile-from-spec when you want to deeply verify a pipeline captures everything from the original spec. Use when prompts might be missing critical domain rules, data models, or invariants.

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。

harperreed/dotfiles3342026年10月3日 更新

harperreed のスキルをすべて見る

このスキルの問題を報告する