SuperAPI AI Directory

create-wiki

General & other

View repository

Updated 10-16 09:32

About this skill

Create Wiki

Critical rules

  • Write reusable technology knowledge, never documentation for the containing project. Exclude project architecture, requirements, decisions, status, and project-specific examples.
  • Store each technology or topic under wiki/<topic-name>/ with descriptive topic-area files such as best-practices.md or queues-and-jobs.md.
  • Do not create README.md, index, overview-only, or navigation files. Filenames and focused content must make discovery sufficient.
  • Research before writing. Prefer official, version-matched documentation, source, changelogs, standards, and first-party examples; corroborate risky or ambiguous claims.
  • Synthesize instead of copying. Retain source URLs while making each file useful without reopening every source.
  • Ask the user when material scope, audience, version, or topic-priority ambiguity would change the wiki.

Content standard

Each file should provide only the sections useful for its topic:

  • concise mental model and key takeaways
  • correct, runnable examples with assumptions and version constraints
  • day-to-day workflows and recommended defaults
  • common failure modes, edge cases, anti-patterns, and diagnostics
  • trade-offs and when an alternative is preferable
  • links to authoritative sources

Prefer several focused files over one exhaustive document. Avoid generic introductions, duplicated documentation, unsupported opinion, and content saved “just in case.”

Workflow

  1. Frame. Confirm the technology, versions, audience, depth, and requested output. Separate reusable knowledge from project-specific material; route the latter to project documentation, create-plan, or another owning workflow.
  2. Inventory. Read the existing wiki/<topic-name>/ files and relevant dependency/configuration evidence. Identify stale claims, gaps, overlap, and established style before choosing filenames.
  3. Research. Investigate official documentation, implementation/source, changelogs, and credible real-world failure evidence. Resolve version conflicts; where sources disagree or version behavior can't be established, label the uncertainty and its practical consequence rather than guessing.
  4. Design. Choose focused topic-area files around actual tasks, boundaries, and pain points. Update existing files instead of creating parallel coverage; preserve accurate material, replace superseded guidance explicitly so no files contradict each other, and leave unrelated topics alone.
  5. Write. Produce concise, actionable Markdown with examples, pitfalls, version notes, and direct source links. Keep all prose and examples portable to another project using the same technology.
  6. Validate. Re-read the whole topic set to remove duplication and verify code/API claims against the targeted version.