Skip to content

AI Skill Index

This page is written as an index for a future AI skill. It is not the skill itself. Use it as the stable map of concepts, files, commands, and docs pages an agent should read before acting in a Cohesive Code workspace.

A Cohesive Code workspace usually has:

  • ccode.yaml at or near the project root
  • a ccode_path directory, defaulting to ccode
  • generated support files under <ccode_path>/.ccode/lib/
  • optional accelerator state under <hidden_path>/accelerators/

If ccode.yaml exists, read it before editing files.

Use this order when references disagree:

  1. The generated local Context type at <ccode_path>/.ccode/lib/context.ts
  2. The current CLI implementation and command output
  3. This documentation
  4. Older README examples

For initialization:

For authoring a process:

For OpenAPI generation:

For accelerated artifacts:

  • Treat ccode.yaml as the workspace contract.
  • Keep source files, templates, specs, and instructions under ccode_path.
  • Do not hand-edit .ccode/build/.
  • Do not edit .ccode/lib/ in an application workspace; rerun ccode init to refresh it.
  • Run the exact process you changed when validation is possible.
  • Use ctx.generate(...) only for files that should be regenerated.
  • Use ctx.accelerate(...) for files that may need human or agent adjustment.
  • Inspect unresolved accelerator work with ccode list accelerated --for-agent.
  • Fetch one adjustment bundle with ccode get accelerated <scopeId>:<artifactId> --instructions --for-agent.
Terminal window
ccode init [path] --version <version>
ccode run <process>
ccode list accelerated [scopeId]
ccode list accelerated [scopeId] --include-resolved
ccode list instructions
ccode list instructions --include-resolved
ccode get accelerated <scopeId>:<artifactId>
ccode get accelerated <scopeId>:<artifactId> --instructions
ccode get instruction <path>

Add --for-agent to list and get commands when the caller needs machine-readable JSON.

ctx.println(message)
ctx.setScope(scopeName)
ctx.scope()
ctx.renderTemplate(templatePath, data)
ctx.generate(templatePath, filePath, data)
ctx.accelerate(id, templatePath, data, instructionsPath?)
ctx.parseJSONFromBytes(jsonBytes)
ctx.parseJSONFromString(jsonString)
ctx.parseJSONFromFile(filePath)
ctx.parseOpenAPIFromBytes(specBytes)
ctx.parseOpenAPIFromString(spec)
ctx.parseOpenAPIFromFile(filePath)