Expand documentation for compile command and CAS features - #1450

Reviewed by checking out the branch and running the things it documents. The fjs compile section is excellent — I ran the example verbatim and the output matches the README byte for byte, both formats. All seven new documentation links resolve. Two problems with the alias→full-name change itself.

1. npx fjs <command> runs somebody else's package

README.md:31:

or run the CLI without installing it, with npx fjs <command>.

npx resolves by package name, not bin name. fjs is a real, unrelated package on npm:

$ npm view fjs name version description repository.url
name = 'fjs'
version = '0.5.0'
description = 'Utilities for Functional Javascript'
repository.url = 'git@github.com:idottv/fjs.git'

So npx fjs test downloads and executes that stranger's package rather than FunctionalScript. Every other npx line in the repo gets this right — AGENTS.md:70, fjs/cas/evo/README.md:242, fjs/mcp/README.md all say npx functionalscript .... Suggest the same here.

The same bug already exists at fjs/README.md:12 (npx fjs <command> [args]) — it predates this PR, but since this PR is precisely about how the CLI is invoked in the docs, it seems like the right place to fix both.

2. Docs and the CLI now disagree about aliases

The commit is titled "Use full command names in documentation, drop CLI aliases", but the aliases are not dropped. fjs/module.f.ts still declares them:

names: ['test', 't'] / ['compile', 'c'] / ['cas', 's'] / ['mcp', 'm'] / ['ci', 'i'] / ['run', 'r']

and node ./fjs/module.ts t still runs the full suite (2354 passing). Meanwhile fjs/README.md drops the Alias column entirely, and README.md adds "Run fjs help for the same list" — but fjs help on this branch prints:

Available commands:
  test, t     Run the FunctionalScript test suite
  compile, c  Compile a FunctionalScript module to JavaScript
  ...
  help, h, ?  Print this help message

That is not the same list. Either drop the aliases in fjs/module.f.ts / fjs/cli/module.f.ts so the help output matches, or keep documenting them. As it stands fjs help is a live counter-example to the README's own claim.

Happy to approve once the npx line is corrected; the alias question is your call on which direction to take it.

Re-checked at a9377224. The update since my last review is the merge from main — it's clean, no conflict markers, and #1449's new JSDoc/variance section in AGENTS.md survived intact. The documentation content itself is unchanged, so both earlier points are still open:

1. npx fjs <command> — still at README.md:31 and fjs/README.md:12. npx resolves by package name, and fjs is a real unrelated package on npm (fjs@0.5.0, "Utilities for Functional Javascript"), so this line tells users to run someone else's code. This is the one I'd call blocking; it's a one-word fix to npx functionalscript <command>.
2. Aliases — fjs/module.f.ts still declares ['test','t'], ['compile','c'], etc., and fjs help still prints them, while the tables no longer do and README.md says fjs help gives "the same list".

No new problems. Everything else I verified before still holds on this head.

Approving at c3ff1173. Both points are fixed, and the second one you took further than I suggested.

npx — now npx functionalscript <command> in both places, and fjs/README.md:16 explains why rather than just changing the string:

fjs is a different package on npm, so npx fjs … would run somebody else's code.

That note is the part that keeps it from regressing.

Aliases — instead of picking one side, you removed the false claim and made the docs state the actual arrangement: aliases exist, fjs help prints them, the documentation uses full names throughout. That's more accurate than either option I offered.

ci: generate full command names — this was the right root-cause fix and it's the one I checked hardest, since it changes a generator rather than prose. Verified:

• npm run ci-update regenerates .github/workflows/ci.yml and leaves the working tree clean — the committed workflow is exactly what the generator produces, so generated CI and docs can't drift apart.
• npx tsc clean; npm test → pass: 2354, fail: 0, total: 2354.
• No t-alias remnants left in the generated workflow or deno.json.

Catching that fjs/ci/README.md was documenting fjs t only because it was quoting real generator output — and fixing the generator instead of the quote — is the correct direction.