agentleFS
WilliamEbongWilliamEbong/CourseForge/CLAUDE.md

CourseForge

CLAUDE.md0 starsChanged 55 days agoMITView on GitHub

CourseForge is a TypeScript/Node course-engineering harness: a deterministic 11-stage pipeline (concept → research → design → storyboard → editorial → visual direction → model → build → QA → release) that uses Claude Code or Codex for semantic work and code for routing, validation, rendering, QA and release decisions. Keep this file consistent with AGENTS.md. 1. Provider-neutral core: only src/harness/** names the claude/codex CLIs, flags or output formats. 2. Only src/core/proc.ts imports child_process (cross-spawn, shell: false, prompts on stdin). 3.…

The file

We split this CLAUDE.md into the rules it gives an agent

5 rules. About 0 are common, 2 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Keep this file consistent with AGENTS.md.

    line 6
  2. Very specific:

    Only src/core/proc.ts imports child_process (cross-spawn, shell: false, prompts on stdin).

    line 48
  3. Very specific:

    No global dependencies: never read or modify ~/.claude, ~/.codex or global npm config.

    line 56
  4. Distinctive:

    No secrets, tokens, personal paths or emails in code, fixtures, logs or docs.

    line 57
  5. Very specific:

    Do not run npm install, commit, stash or check out unless asked.

    line 64

≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
CLAUDE.md

CourseForge: instructions for Claude Code

CourseForge is a TypeScript/Node course-engineering harness: a deterministic 11-stage pipeline (concept → research → design → storyboard → editorial → visual direction → model → build → QA → release) that uses Claude Code or Codex for semantic work and code for routing, validation, rendering, QA and release decisions. Keep this file consistent with AGENTS.md.

Sources of truth

  • docs/exec-plans/active/courseforge-v1.md: the approved plan (architecture, decisions, test mapping)
  • docs/architecture/: how the system works as built; docs/adr/: recorded deviations
  • docs/bootstrap/: original requirements of record (read-only; do not treat as runtime code)
  • config/*.json: registries (stages, reviewers, skills, tools, routing, fallbacks, review policy)
  • src/core/schemas/: zod schemas for every artifact and agent output
  • .claude/rules/: path-scoped rules (architecture.md for src/, courses.md for courses/, testing.md)

Map

PathContents
src/cli, src/environmentCLI (help.ts lists commands and exit codes), doctor/setup
src/pipelineState machine, stage loop (run-stage.ts), stage handlers, write audit, public API (api.ts)
src/routingPure router: execution plans, backend selection, gates, visual routes
src/harnessClaude/Codex/fake adapters, prompt assembly, recording (only place that knows the CLIs)
src/review, src/releaseFindings, adjudication, repair plans, scoped repair; release gate and reports
src/ingestion, src/artifactsAny-stage import; artifact registry, versions, locks, trace graph
src/graphics, src/renderer, components/course-uiRenderers, single-file HTML compiler, UI components and runtime
src/qaPlaywright contract runner, crawler, axe, regression
prompts/templates, prompts/rubrics, .claude/skillsPrompt data (never embed prompt prose in TypeScript)
courses/<id>/One self-contained course project each

Commands

npm run build          # tsc → dist/
npm run typecheck
npm run lint           # Biome; npm run format to fix
npm test               # unit + integration (offline, fake harness)
npm run test:e2e       # Playwright
npm run test:acceptance
npm run gen:schemas    # after changing src/core/schemas
node bin/courseforge.mjs <command>    # or ./courseforge, .\courseforge

Non-negotiables

  1. Provider-neutral core: only src/harness/** names the claude/codex CLIs, flags or output formats.
  2. Only src/core/proc.ts imports child_process (cross-spawn, shell: false, prompts on stdin).
  3. Deterministic routing: stages, reviewers, skills, tools, renderers and fallbacks come from config/ via src/routing. Agents return closed enums; code maps them to tools. Plans are saved before agent work.
  4. zod schemas are the source of truth. After changing one, run npm run gen:schemas and commit schemas/. Agent-facing schemas stay flat and strict (all required, optional = nullable, no $ref).
  5. Course isolation: course reads/writes stay under courses/<id>/, through src/core/fsx.ts helpers.
  6. Preserve human work: never edit input/originals/, versions/, locked artifacts or locked IDs; agent steps run inside the write audit, which must not be bypassed.
  7. No global dependencies: never read or modify ~/.claude, ~/.codex or global npm config; no global installs.
  8. No secrets, tokens, personal paths or emails in code, fixtures, logs or docs.

Parallel work

  • Work only in the directories you were assigned; do not edit other streams' files.
  • The coordinator alone edits package.json, the lockfile, tsconfig*.json, biome.json, test runner configs and shared interfaces (src/core/*, src/harness/types.ts, src/pipeline/api.ts).
  • Do not run npm install, commit, stash or check out unless asked. Do not run e2e while other browser or agent-pool work is running (limited RAM).

Before you say you are done

  • npm run typecheck, npm run lint, npm test pass.
  • npm run test:e2e passes if you touched the renderer, components, runtime, graphics or QA.
  • Schema changes: npx tsx scripts/gen-schemas.ts --check passes.
  • New or changed tests covering a requirement carry its acceptance tag (@C6 …).
  • Docs in docs/ still describe the code; UI changes were checked in a real browser.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.