Thanks for helping improve Agent Wallclock. This project is local-first temporal context for language models — contributions should preserve that honesty.
- Read
README.mdfor product scope (not a hosted chat UI). - Run the no-leak checklist before pushing.
- For pass-batch work, follow
docs/PASS_PR_NAMING.md.
git clone https://github.com/TelivityAI/agent-wallclock.git
cd agent-wallclock
npm install
npm run build
npm link -w @agent-wallclock/cliVerify:
wallclock now
wallclock doctor| Type | Pattern | Example |
|---|---|---|
| Pass / hardening | fix/pass-NNN-slug |
fix/pass-hardening-100 |
| Feature | feat/short-slug |
feat/timeline-json |
| Docs only | docs/short-slug |
docs/mcp-readme |
| Chore | chore/short-slug |
chore/ci-node-22 |
Every PR that changes behavior must pass:
npm test # unit tests (@agent-wallclock/core)
npm run smoke # build + CLI smoke + MCP smokeCI runs the same on Ubuntu with Node 22. Fix failures before requesting review.
- Fill out the PR template.
- One logical change per PR when possible.
- Update adapters when
MODEL_RULESor briefing shape changes (adapters/*,catalog/models.md). - Docs must stay honest: no fake product screenshots; label illustrative UI as MOCK.
- Use path placeholders (
/ABSOLUTE/PATH/TO/...) — never commit home directories or tokens.
- TypeScript in
packages/*, ESM, Node 20+ - Match existing naming and error handling (
CliError,ExitCode) - Minimize scope — focused diffs over drive-by refactors
packages/core/src/brief.ts exports MODEL_RULES. Host adapters should stay aligned on:
- Trust briefing only while fresh (Generated at / Stale after, default 15m)
- Never invent time, session age, or effort duration
- Missing fields → unknown
Maintainers use docs/SHIP_CHECKLIST.md before tagging. Contributors do not need to cut releases unless asked.
Open a feature request or discussion issue. For vulnerabilities, see SECURITY.md.