Skip to content
T1KTheOneKit

t1k:cocos:playable:extract

FieldValue
Moduleplayable
Version0.5.6
Efforthigh
Tools—

Keywords: analyze, bundle, competitor, decompile, extract, playable, playable ad, reverse engineer

How to invoke

Section titled “How to invoke”
/t1k:cocos:playable:extract
<url>

Cocos Playable Extract — Competitor Playable Reverse-Engineer

Section titled “Cocos Playable Extract — Competitor Playable Reverse-Engineer”

Given a competitor playable-ad HTML bundle URL, produce 3 commit-ready English docs and a gitignored raw/ artifact tree.

Usage

Section titled “Usage”
/t1k:cocos:playable:extract <https-url-to-playable.html>

Pipeline (5 runtime phases)

Section titled “Pipeline (5 runtime phases)”
  1. Fetch — scripts/fetch-bundle.cjs <url> <outDir>/raw/ — crawls same-origin HTML/JS/img/audio/json, writes raw/manifest.json (sourceUrl, timestamp, fileList, sha256Map, cacheKey). Hard cap: 10 MB.
  2. Unpack — scripts/unpack-bundle.cjs <outDir>/raw/ — extracts inline <script> tags, decodes base64, beautifies JS via js-beautify.
  3. Detect — engine fingerprint (cocos creator 3.x / 2.x / pixi / playcanvas / custom) using regex catalog in references/engine-fingerprints.md. Output: detected engine + version + confidence + entry-point JS file.
  4. Analyze — spawn t1k-cocos-playable-extractor subagent (background, fork context). Subagent reads beautified JS + assets, infers gameplay loop, code architecture, monetization hooks (catalog: references/monetization-hooks.md).
  5. Generate — subagent writes 3 docs from templates in references/templates/:
    • gameplay.md — genre, core loop, win/lose, CTAs, session length, monetization
    • code-structure.md — engine + version, bundle breakdown, entry point, modules, event flow, obfuscation level
    • cocos-implementation.md — Cocos rebuild recipe; MUST cite ≥5 distinct t1k-cocos-playable-* skills with concrete invocation arguments

Skill Body Steps (when invoked)

Section titled “Skill Body Steps (when invoked)”
  1. Validate <url> is http(s); reject otherwise
  2. Compute slug = sha256(url).slice(0,8) + '-' + guessGameNameFromUrl(url) (rules: references/slug-rules.md)
  3. Resolve outDir = plans/research/playable-{slug}/
  4. Cache check: if outDir/raw/manifest.json matches input URL hash → skip to step 7
  5. Run scripts/fetch-bundle.cjs → smoke first file, surface result, gate via user confirm before full crawl (per rules/preview-first-batch.md)
  6. Run scripts/unpack-bundle.cjs
  7. Spawn t1k-cocos-playable-extractor subagent in background with arguments: rawDir, module-inventory (from metadata.json.installedModules + t1k-activation-*.json), 3 doc output paths
  8. Skill validates output: each doc exists, cocos-implementation.md cites ≥5 distinct t1k-cocos-playable-* skills, no inline binary, no PII (URLs/tokens/query params stripped via scripts/lib/sanitize.cjs)
  9. Append outDir/raw/ to .gitignore under marker block # t1k:cocos:playable:extract BEGIN/END (append-only)
  10. Commit outDir/raw/manifest.json only — full raw assets stay gitignored (Q2=B resolution)
  11. Final prompt: “Run /t1k:plan against cocos-implementation.md?”

Output Layout

Section titled “Output Layout”
plans/research/playable-{slug}/
├── gameplay.md                  # committed
├── code-structure.md            # committed
├── cocos-implementation.md      # committed
└── raw/
    ├── manifest.json            # committed (Q2=B)
    ├── index.html               # gitignored
    ├── scripts/*.js             # gitignored
    ├── scripts/*.beautified.js  # gitignored
    └── assets/                  # gitignored

Constraints

Section titled “Constraints”
  • Static analysis only — no headless execution, no source-map de-obfuscation
  • 10 MB hard cap on total bundle size; refuses past cap with explicit message (Q3=A resolution)
  • Single URL per invocation — no batch mode
  • Cocos rebuild recipe always emitted, regardless of detected engine (kit’s purpose)
  • Confidence tagging mandatory — every doc footer states confidence (high/med/low) with reason; subagent must fall back to “behavioral spec” mode rather than fabricate identifiers when obfuscation defeats analysis

Common Mistakes

Section titled “Common Mistakes”
  • Hardcoding ad-network CDN domains in scripts — keep configurable via input URL only
  • Re-fetching when cache is valid — always check manifest.json first
  • Committing raw/ assets — only manifest.json is committed; full tree is gitignored
  • Citing fewer than 5 Cocos skills in cocos-implementation.md — validator must reject

Gotchas

Section titled “Gotchas”
  • Mintegral CDN 401/403 — some assets walled; fetch-bundle.cjs logs gap to manifest and continues with available files. Subagent must flag “partial bundle — analysis may miss assets” in code-structure.md.
  • Obfuscated identifiers — t/e/n minified variable names defeat structure analysis. Subagent uses “describe-by-behavior” mode: notes obfuscation: heavy — identifiers unrecoverable instead of fabricating names.
  • Cocos runtime is ~2 MB of boilerplate — subagent must skip engine code (regex-detected via references/engine-fingerprints.md) to preserve token budget for game-specific code.
  • URL sanitizer is SSOT — never write a raw URL to manifest, docs, or chat; route every URL through scripts/lib/sanitize.cjs first to strip tracking params and PII.
Section titled “Related”
  • t1k-cocos-playable-build-size — bundle-size targets the recipe references
  • t1k-cocos-playable-fsm, t1k-cocos-playable-signalbus, t1k-cocos-playable-lifecycle, t1k-cocos-playable-object-pool, t1k-cocos-playable-parameter — core skills the rebuild recipe cites
  • rules/preview-first-batch.md — smoke-first gate enforced in Step 5
  • rules/code-conventions-cocos.md — file scaffold the recipe matches