@pigcloud/skills 1.0.10 → 1.1.2

package/skills/business-fact-extraction/SKILL.md

@@ -1,337 +0,0 @@
----
-name: business-fact-extraction
-description: 当 Codex 需要逆向理解系统真实行为或为领域建模、需求细化或知识沉淀准备证据时，从代码、测试、日志和文档中提取业务事实、当前行为、不变量和术语
-lifecycle_stage: analysis
-rule_profile: modeling
-dependencies: []
-triggers:
-  - reverse engineer behavior
-  - business facts
-  - current behavior
-  - code facts
-  - extract facts from code
-  - behavior map
-inputs:
-  - code paths
-  - tests
-  - logs
-  - docs
-  - schemas
-  - runtime evidence
-outputs:
-  - code fact map
-  - business fact map
-  - codewiki delta
-  - handoff notes
-  - knowledge-base file targets
-workflow:
-  - collect code, test, doc, and runtime evidence for the requested area
-  - extract code facts first, then derive business meaning, then propose CodeWiki updates
-  - mark each fact as evidence-backed, inferred, or unknown
-  - summarize gaps, contradictions, and open questions
-  - choose the destination knowledge-base files before writing the final note
-  - hand off to domain-modeling, spec-refinement, or knowledge-capture
-gates:
-  - stop at fact extraction
-  - do not design a new solution
-  - do not rewrite the PRD
-  - do not generate code
-refs:
-  - skills/references/business-fact-extraction.md
-  - skills/references/skill-reference-matrix.md
-  - skills/references/rule-loading-map.md
-  - docs/codewiki-knowledge-base.md
----
-
-# Business Fact Extraction
-
-## Purpose
-
-Turn code-backed evidence into reusable business facts.
-
-## Suitable / Unsuitable
-
-- Suitable: reverse engineering, current behavior mapping, business rule extraction, terminology normalization, evidence-backed summaries
-- Unsuitable: PRD drafting, solution design, implementation, review, final knowledge capture
-
-## Quick Start
-
-- Start from code paths, tests, docs, runtime traces, and schemas.
-- Produce the output in three stages: code facts, business facts, and CodeWiki updates.
-- Keep evidence, inference, and unknowns distinct.
-- Hand off the result when the facts are stable enough for `domain-modeling`, `spec-refinement`, or `knowledge-capture`.
-
-## Write Back
-
-- When the user wants persistence, render the staged output into a JSON fact pack.
-- Use `skills/business-fact-extraction/scripts/write-knowledge-base.js` to write the pack into the knowledge-base files.
-- Keep the pack keyed by target file path under `knowledge-base/`.
-- Write only the smallest matching file for each fact.
-- Treat the script as a writer, not as a second analysis pass.
-
-## Inputs / Outputs
-
-- Inputs:
-  - code paths
-  - tests
-  - logs
-  - docs
-  - schemas
-  - runtime evidence
-- Outputs:
-  - code fact map
-  - business fact map
-  - codewiki delta
-  - handoff notes
-  - knowledge-base file targets
-
-## Output Template
-
-### 1. Code Facts
-
-#### `knowledge-base/realtime-code/modules.md`
-
-- module:
-- current responsibility:
-- key entry points:
-- important constraints:
-- last verified:
-
-#### `knowledge-base/realtime-code/flows.md`
-
-- flow name:
-- trigger:
-- path:
-- output:
-- failure points:
-- notes:
-
-#### `knowledge-base/realtime-code/runtime-notes.md`
-
-- fact:
-- evidence:
-- affected area:
-- date:
-- notes:
-
-### 2. Business Facts
-
-#### `knowledge-base/requirements/decisions.md`
-
-- decision:
-- reason:
-- date or context:
-- affected artifact:
-- notes:
-
-#### `knowledge-base/requirements/acceptance.md`
-
-- requirement:
-- acceptance criteria:
-- out of scope:
-- open questions:
-- confirmed by:
-
-#### `knowledge-base/requirements/conflicts.md`
-
-- conflict:
-- source:
-- impact:
-- resolution:
-- status:
-
-### 3. CodeWiki Update
-
-#### `knowledge-base/codewiki/overview.md`
-
-- one-line summary:
-- what the system does:
-- what changed:
-- why it matters:
-
-#### `knowledge-base/codewiki/architecture.md`
-
-- layers:
-- boundaries:
-- dependencies:
-- major integration points:
-
-#### `knowledge-base/codewiki/modules.md`
-
-- module name:
-- responsibility:
-- main interfaces:
-- related runtime facts:
-
-#### `knowledge-base/codewiki/flows.md`
-
-- request flow:
-- data flow:
-- error flow:
-- handoff flow:
-
-#### `knowledge-base/codewiki/api.md`
-
-- endpoint or contract:
-- request shape:
-- response shape:
-- constraints:
-
-#### `knowledge-base/codewiki/runbook.md`
-
-- startup:
-- verify:
-- troubleshoot:
-- recover:
-
-#### `knowledge-base/codewiki/conventions.md`
-
-- naming:
-- layering:
-- error handling:
-- logging:
-- data handling:
-
-### 4. Enterprise Note
-
-#### `knowledge-base/enterprise/decisions.md`
-
-- decision:
-- reason:
-- scope:
-- date:
-- reusable lesson:
-
-#### `knowledge-base/enterprise/pitfalls.md`
-
-- pitfall:
-- context:
-- why it failed:
-- how to avoid it:
-
-#### `knowledge-base/enterprise/examples.md`
-
-- example:
-- context:
-- lesson:
-- reusable pattern:
-
-#### `knowledge-base/enterprise/ai-coding-rules.md`
-
-- rule:
-- when it applies:
-- why it matters:
-- examples:
-
-## Sample Output
-
-### Code Facts
-
-#### `knowledge-base/realtime-code/modules.md`
-
-- module: `order-service`
-- current responsibility: creates orders and reserves inventory
-- key entry points: `createOrder()`, `reserveInventory()`
-- important constraints: stock must be checked before commit
-- last verified: 2026-06-22
-
-#### `knowledge-base/realtime-code/flows.md`
-
-- flow name: checkout order placement
-- trigger: user confirms checkout
-- path: validate stock -> reserve inventory -> create order
-- output: confirmed order or shortage error
-- failure points: stock shortage, reservation failure
-- notes: reservation must happen before the order commit
-
-### Business Facts
-
-#### `knowledge-base/requirements/decisions.md`
-
-- decision: inventory must be reserved before order confirmation
-- reason: avoid overselling
-- date or context: 2026-06-22, order flow review
-- affected artifact: checkout spec-refinement
-- notes: failure should return a clear stock shortage message
-
-#### `knowledge-base/requirements/acceptance.md`
-
-- requirement: checkout must not confirm an order when stock is unavailable
-- acceptance criteria: shortage blocks confirmation and returns a clear message
-- out of scope: backorder design
-- open questions: whether to retry reservation automatically
-- confirmed by: order flow review
-
-### CodeWiki Update
-
-#### `knowledge-base/codewiki/flows.md`
-
-- request flow: checkout -> validate stock -> reserve inventory -> create order
-- data flow: cart snapshot -> reservation record -> order record
-- error flow: stock shortage -> stop checkout -> return shortage reason
-- handoff flow: checkout service hands off to inventory service and order repository
-
-#### `knowledge-base/codewiki/modules.md`
-
-- module name: `order-service`
-- responsibility: coordinates checkout and order creation
-- main interfaces: `createOrder()`, `reserveInventory()`
-- related runtime facts: inventory must be reserved before commit
-
-### Enterprise Note
-
-#### `knowledge-base/enterprise/pitfalls.md`
-
-- pitfall: reservation happens after order creation
-- context: checkout path
-- why it failed: allowed overselling under concurrent requests
-- how to avoid it: reserve before committing the order
-
-#### `knowledge-base/enterprise/ai-coding-rules.md`
-
-- rule: reserve inventory before committing the order
-- when it applies: any checkout flow that can oversell
-- why it matters: keeps business state consistent under concurrency
-- examples: validate stock, reserve, then create the order
-
-## Workflow
-
-1. Collect code, test, doc, and runtime evidence for the requested area.
-2. Write the code fact map with observed behavior, state changes, contracts, and evidence.
-3. Derive the business fact map from the code facts and separate inference from proof.
-4. Draft the CodeWiki delta with reusable architecture, flow, and convention updates.
-5. Map each section to the exact knowledge-base files it should update.
-6. Mark each fact as evidence-backed, inferred, or unknown.
-7. Summarize gaps, contradictions, and open questions.
-8. Hand off to `domain-modeling`, `spec-refinement`, or `knowledge-capture`.
-
-## Replay Signals
-
-- Input signal: codebase walkthrough, behavior reconstruction, business rule extraction, or current-state analysis.
-- Output to verify: code fact map, business fact map, codewiki delta, file targets, gap list.
-- Stop signal: solution design, PRD rewriting, or code generation.
-- Handoff signal: the extracted facts are stable enough for downstream modeling or capture.
-
-## Gotchas
-
-- Do not infer business intent from comments alone.
-- Do not merge evidence-backed facts and assumptions into one claim.
-- Do not skip contradictory evidence.
-- Do not turn extraction into design.
-- Do not write the CodeWiki delta before the code facts are stable.
-- Do not leave file targets implicit.
-
-## Stop Rules
-
-- Stop at fact extraction.
-- Do not design a new solution.
-- Do not rewrite the PRD.
-- Do not generate code.
-
-## References
-
-- `skills/references/business-fact-extraction.md`
-- `skills/references/skill-reference-matrix.md`
-- `skills/references/rule-loading-map.md`
-- `docs/codewiki-knowledge-base.md`
-

package/skills/business-fact-extraction/scripts/write-knowledge-base.js

@@ -1,228 +0,0 @@
-#!/usr/bin/env node
-
-const fs = require('fs');
-const path = require('path');
-
-const REPO_ROOT = path.join(__dirname, '..', '..', '..');
-const START_MARKER = '<!-- business-fact-extraction:start -->';
-const END_MARKER = '<!-- business-fact-extraction:end -->';
-
-const FILE_FIELD_ORDER = {
-  'knowledge-base/realtime-code/modules.md': ['module', 'current responsibility', 'key entry points', 'important constraints', 'last verified'],
-  'knowledge-base/realtime-code/flows.md': ['flow name', 'trigger', 'path', 'output', 'failure points', 'notes'],
-  'knowledge-base/realtime-code/runtime-notes.md': ['fact', 'evidence', 'affected area', 'date', 'notes'],
-  'knowledge-base/requirements/decisions.md': ['decision', 'reason', 'date or context', 'affected artifact', 'notes'],
-  'knowledge-base/requirements/acceptance.md': ['requirement', 'acceptance criteria', 'out of scope', 'open questions', 'confirmed by'],
-  'knowledge-base/requirements/conflicts.md': ['conflict', 'source', 'impact', 'resolution', 'status'],
-  'knowledge-base/codewiki/overview.md': ['one-line summary', 'what the system does', 'what changed', 'why it matters'],
-  'knowledge-base/codewiki/architecture.md': ['layers', 'boundaries', 'dependencies', 'major integration points'],
-  'knowledge-base/codewiki/modules.md': ['module name', 'responsibility', 'main interfaces', 'related runtime facts'],
-  'knowledge-base/codewiki/flows.md': ['request flow', 'data flow', 'error flow', 'handoff flow'],
-  'knowledge-base/codewiki/api.md': ['endpoint or contract', 'request shape', 'response shape', 'constraints'],
-  'knowledge-base/codewiki/runbook.md': ['startup', 'verify', 'troubleshoot', 'recover'],
-  'knowledge-base/codewiki/conventions.md': ['naming', 'layering', 'error handling', 'logging', 'data handling'],
-  'knowledge-base/enterprise/decisions.md': ['decision', 'reason', 'scope', 'date', 'reusable lesson'],
-  'knowledge-base/enterprise/pitfalls.md': ['pitfall', 'context', 'why it failed', 'how to avoid it'],
-  'knowledge-base/enterprise/examples.md': ['example', 'context', 'lesson', 'reusable pattern'],
-  'knowledge-base/enterprise/ai-coding-rules.md': ['rule', 'when it applies', 'why it matters', 'examples']
-};
-
-const SAMPLE_PACK = {
-  files: {
-    'knowledge-base/realtime-code/modules.md': [
-      {
-        module: 'order-service',
-        'current responsibility': 'creates orders and reserves inventory',
-        'key entry points': 'createOrder(), reserveInventory()',
-        'important constraints': 'stock must be checked before commit',
-        'last verified': '2026-06-22'
-      }
-    ],
-    'knowledge-base/requirements/decisions.md': [
-      {
-        decision: 'inventory must be reserved before order confirmation',
-        reason: 'avoid overselling',
-        'date or context': '2026-06-22, order flow review',
-        'affected artifact': 'checkout spec-refinement',
-        notes: 'failure should return a clear stock shortage message'
-      }
-    ],
-    'knowledge-base/codewiki/flows.md': [
-      {
-        'request flow': 'checkout -> validate stock -> reserve inventory -> create order',
-        'data flow': 'cart snapshot -> reservation record -> order record',
-        'error flow': 'stock shortage -> stop checkout -> return shortage reason',
-        'handoff flow': 'checkout service hands off to inventory service and order repository'
-      }
-    ],
-    'knowledge-base/enterprise/pitfalls.md': [
-      {
-        pitfall: 'reservation happens after order creation',
-        context: 'checkout path',
-        'why it failed': 'allowed overselling under concurrent requests',
-        'how to avoid it': 'reserve before committing the order'
-      }
-    ]
-  }
-};
-
-function parseArgs(argv) {
-  const args = {
-    input: null,
-    root: REPO_ROOT,
-    dryRun: false,
-    sample: false
-  };
-
-  for (let index = 0; index < argv.length; index += 1) {
-    const token = argv[index];
-    if (token === '--input' || token === '-i') {
-      args.input = argv[++index];
-      continue;
-    }
-    if (token.startsWith('--input=')) {
-      args.input = token.slice('--input='.length);
-      continue;
-    }
-    if (token === '--root' || token === '-r') {
-      args.root = path.resolve(argv[++index]);
-      continue;
-    }
-    if (token.startsWith('--root=')) {
-      args.root = path.resolve(token.slice('--root='.length));
-      continue;
-    }
-    if (token === '--dry-run') {
-      args.dryRun = true;
-      continue;
-    }
-    if (token === '--sample') {
-      args.sample = true;
-      continue;
-    }
-  }
-
-  return args;
-}
-
-function readInput(inputPath) {
-  if (!inputPath || inputPath === '-') {
-    return fs.readFileSync(0, 'utf8');
-  }
-  return fs.readFileSync(inputPath, 'utf8');
-}
-
-function loadPack(inputText) {
-  const parsed = JSON.parse(String(inputText).replace(/^\uFEFF/, ''));
-  if (parsed && typeof parsed === 'object' && parsed.files && typeof parsed.files === 'object') {
-    return parsed.files;
-  }
-  if (parsed && typeof parsed === 'object') {
-    return parsed;
-  }
-  throw new Error('Input must be a JSON object.');
-}
-
-function normalizeValue(value) {
-  if (value === null || value === undefined || value === '') return 'unknown';
-  if (Array.isArray(value)) return value.map(item => normalizeValue(item)).join(', ');
-  if (typeof value === 'object') return JSON.stringify(value);
-  return String(value).trim() || 'unknown';
-}
-
-function renderEntry(fields, entry) {
-  return fields.map(field => `- ${field}: ${normalizeValue(entry[field])}`).join('\n');
-}
-
-function renderFileSection(relativePath, entries) {
-  const fieldOrder = FILE_FIELD_ORDER[relativePath];
-  if (!fieldOrder) {
-    throw new Error(`Unsupported target file: ${relativePath}`);
-  }
-  if (!Array.isArray(entries) || entries.length === 0) {
-    throw new Error(`No entries provided for ${relativePath}`);
-  }
-
-  const blocks = entries.map(entry => renderEntry(fieldOrder, entry));
-  return [
-    '## Generated Entries',
-    START_MARKER,
-    blocks.join('\n\n'),
-    END_MARKER
-  ].join('\n');
-}
-
-function upsertGeneratedSection(existingContent, generatedSection) {
-  const startIndex = existingContent.indexOf(START_MARKER);
-  const endIndex = existingContent.indexOf(END_MARKER);
-
-  if (startIndex >= 0 && endIndex >= 0 && endIndex > startIndex) {
-    const before = existingContent.slice(0, startIndex).trimEnd();
-    const after = existingContent.slice(endIndex + END_MARKER.length).trimStart();
-    const middle = generatedSection;
-    const parts = [before, middle];
-    if (after) {
-      parts.push(after);
-    }
-    return `${parts.filter(Boolean).join('\n\n').trimEnd()}\n`;
-  }
-
-  const base = existingContent.trimEnd();
-  if (!base) {
-    return `${generatedSection}\n`;
-  }
-
-  return `${base}\n\n${generatedSection}\n`;
-}
-
-function collectTargets(filePack) {
-  const targets = [];
-  for (const [relativePath, entries] of Object.entries(filePack)) {
-    if (!relativePath.startsWith('knowledge-base/')) {
-      throw new Error(`Refusing to write outside knowledge-base: ${relativePath}`);
-    }
-    targets.push({ relativePath, entries });
-  }
-  return targets;
-}
-
-function ensureParentDir(filePath) {
-  fs.mkdirSync(path.dirname(filePath), { recursive: true });
-}
-
-function main() {
-  const args = parseArgs(process.argv.slice(2));
-
-  if (args.sample) {
-    process.stdout.write(`${JSON.stringify(SAMPLE_PACK, null, 2)}\n`);
-    return;
-  }
-
-  const inputText = readInput(args.input);
-  const filePack = loadPack(inputText);
-  const targets = collectTargets(filePack);
-
-  for (const { relativePath, entries } of targets) {
-    const absolutePath = path.join(args.root, relativePath);
-    const generatedSection = renderFileSection(relativePath, entries);
-    const existingContent = fs.existsSync(absolutePath) ? fs.readFileSync(absolutePath, 'utf8') : '';
-    const nextContent = upsertGeneratedSection(existingContent, generatedSection);
-
-    if (args.dryRun) {
-      process.stdout.write(`=== ${relativePath} ===\n${nextContent}\n`);
-      continue;
-    }
-
-    ensureParentDir(absolutePath);
-    fs.writeFileSync(absolutePath, nextContent, 'utf8');
-    process.stdout.write(`Wrote ${relativePath}\n`);
-  }
-}
-
-try {
-  main();
-} catch (error) {
-  console.error(error.message);
-  process.exitCode = 1;
-}
-

package/skills/code-review/SKILL.md

@@ -1,136 +0,0 @@
----
-name: code-review
-description: 当需要通用代码审查时，对 diff 或 PR 进行正确性、可维护性和回归风险审查
-lifecycle_stage: review
-rule_profile: code
-dependencies:
-  - feature-build
-triggers:
-  - diff
-  - pull request
-  - review
-  - quality gate
-  - checklist
-  - code review
-inputs:
-  - diff
-  - pull request
-  - changed files
-outputs:
-  - findings
-  - severity
-  - fix recommendation
-  - evidence summary
-  - review summary
-workflow:
-  - lock the review scope
-  - identify the tech stack and load the matching review overlay(s)
-  - inspect correctness, maintainability, and regression risk
-  - route security-specific issues to `security-audit`
-  - route performance-specific issues to `performance-audit`
-  - rank findings by severity
-  - produce an actionable review report
-gates:
-  - stop at review findings
-  - do not rewrite code
-  - do not perform implementation refactors
-  - do not become a general design discussion
-refs:
-  - skills/references/rule-loading-map.md
-  - skills/references/flow-test-cases.md
-  - skills/references/full-chain-replay-scenarios.md
-  - skills/references/prompt-replay-checklist.md
-  - skills/code-review/references/rules.md
-  - skills/code-review/references/findings-template.md
-  - skills/code-review/references/stack-notes.md
-  - rules/index.md
----
-
-# Code Review
-
-## Purpose
-
-Review a change set and return actionable findings.
-
-## Suitable / Unsuitable
-
-- Suitable: PR review, diff review, quality gate, regression check, generic code quality review across Java, Vue, and mixed-stack changes
-- Unsuitable: implementation work, refactoring, broad design discussion, security-specific review, performance-specific review, requirement rewriting
-
-## Quick Start
-
-- Lock the review scope before reading details.
-- Identify the stack(s) first. This repository may apply Java, Vue, or mixed-stack review overlays.
-- Read the diff with the current rules bundle in mind.
-- Report findings with severity and a concrete fix path.
-- If the issue is clearly about auth, secrets, or exposure, hand off to `security-audit`.
-- If the issue is clearly about latency, throughput, or resource usage, hand off to `performance-audit`.
-- Keep stack-specific concerns in the matching overlay and keep the core review generic.
-
-## Report Format
-
-- Use one entry per finding.
-- Include `location`, `severity`, `evidence`, `why it matters`, and `fix recommendation`.
-- When a stack overlay applies, anchor the finding to the smallest stack-specific surface:
-  - Java: controller, service, entity, mapper, DTO / VO, or MyBatis XML symbol
-  - Vue: component, composable, state module, template, or frontend build file
-  - TypeScript: typed utility, module boundary, state module, or TS config symbol
-- Keep the review summary short when the change is acceptable.
-- If there are no findings, explain why the change is acceptable and what was checked.
-- Separate must-fix findings from optional suggestions.
-
-## Inputs / Outputs
-
-- Inputs:
-  - diff
-  - pull request
-  - changed files
-- Outputs:
-  - findings
-  - severity
-  - fix recommendation
-  - evidence summary
-
-## Workflow
-
-1. Lock the review scope.
-2. Identify the stack overlays and load the matching review baseline(s) when applicable.
-3. Inspect correctness, maintainability, and regression risk.
-4. Route security-specific issues to `security-audit`.
-5. Route performance-specific issues to `performance-audit`.
-6. Rank findings by severity.
-7. Produce an actionable review report.
-
-## Replay Signals
-
-- Input signal: diff, PR, changed files, or quality gate request.
-- Output to verify: findings, severity, fix recommendation, evidence summary, review summary.
-- Stop signal: rewriting code, implementation refactors, or broad design discussion.
-- Handoff signal: findings are ready for `knowledge-capture`.
-
-## Gotchas
-
-- Do not turn review into implementation work.
-- Do not broaden the scope beyond the submitted change.
-- Do not produce findings without concrete evidence.
-- Do not hide severity or fix recommendation.
-- Do not keep security or performance findings inside generic code review when a specialized review skill fits better.
-- Do not turn stack-specific rules into a generic design discussion.
-
-## Stop Rules
-
-- Stop at review findings.
-- Do not rewrite code.
-- Do not perform implementation refactors.
-- Do not become a general design discussion.
-- Do not turn review into a design task.
-
-## References
-
-- `skills/references/rule-loading-map.md`
-- `skills/references/flow-test-cases.md`
-- `rules/review/index.md`
-- `rules/review/code.md`
-- `rules/review/java.md`
-- `rules/review/vue.md`
-