code-unpacked
Skill1 starsChanged 4 days agoMITView on GitHub
Generate all five Code Unpacked formats for a PR, branch, codebase, or selected code. Use when a complete bundle or all formats are requested.
In this repository5 other files it gives its agents.
- code-mapSkill
- reverse-designSkill
- review-quizSkill
- walkthroughSkill
- xkcdSkill
The file
Files
Code Unpacked
Produce a walkthrough, reverse-design documents, code map, review quiz, and comics from one source snapshot and one shared analysis. Run the workflow in the current agent; no other installed skills or parallel agents are required.
Use
/code-unpacked pr https://github.com/owner/repo/pull/123 --draft
/code-unpacked branch feature/login --base main
/code-unpacked codebase
/code-unpacked docs auth --files src/auth.ts:10-90,src/session.tsAccept natural-language requests and the harness's native invocation syntax. Interpret PR URLs as pr, named branches as branch, selected files or pasted code as named docs, and the whole current repository as codebase. Honor requested quiz length and comic counts/panels. Video is an optional addition only when requested; the normal walkthrough is HTML.
Capture and analyze once
Read the bundled source and delivery guide and shared analysis guide. Resolve <core> to this installed skill's runtime directory. Capture the target once with its collector, inspect the source, author the shared review analysis, and validate/save it through analyze. Reuse an existing capture and analysis only when they match the requested source revision and scope.
Use the same saved context.json, evidence IDs, and analysis for every presentation. Do not repeat collection or analysis when reading the individual workflows below. If shared source capture or analysis cannot be validated, stop and explain the cause. Dependencies and browsers are set up automatically when needed.
Generate all five formats
The following workflows and renderers are bundled here. For each workflow, <skill-dir> means this installed skill's presentations/<name> directory, and <core> remains the shared runtime directory. These instructions override workflow references to capturing source again or looking for sibling skills.
| Format | Authoring workflow |
| Walkthrough | Chapters, source excerpts, and optional video |
| Reverse-design | Relevant design documents and diagrams |
| Code-map | Files, functions, relationships, and flows |
| Review-quiz | Questions, answers, and recommendations |
| Xkcd | Storyboards, generated artwork, and transcript |
Read each workflow as you reach that format, author its content against the shared analysis, and run its existing renderer with the same --context and target --root. Render in the order above. Keep content JSON and artifact generation internal; the user should not need to run scripts or prepare inputs.
For comics, use an available image-generation tool and inspect the complete artwork. Initial rendering without artwork saves prompts and returns an incomplete result; this is an expected preparation step. If image generation is unavailable, preserve those prompts and the storyboard and report comics as incomplete. Continue producing all other formats.
Catch failures per format, preserve completed outputs, and continue with the remaining formats. If requested video cannot be exported, retain the HTML walkthrough and report the video failure separately. Do not label a previous run's artifact as a current success: check the renderer result and its manifest against this run's shared context.
Return and deliver
Return one concise list of the five results with usable artifact links and statuses: complete, partial, incomplete, or failed. Explain missing capabilities briefly. Outputs retain their existing .code-unpacked/<scope>/<format>/ locations and per-format manifests; do not create a new artifact schema or claim incomplete comics are complete.
For PRs, prepare each format's existing draft comment and attachment workflow. Post only when the user requested live delivery, and only post complete, publishable outputs. --draft always stays local. Missing GitHub or browser-upload access preserves local results and is reported separately from generation status.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

