docs: close top-4 documentation gaps across AGENTS, ARCHITECTURE, PRD, tooling #4
Loading…
Reference in a new issue
No description provided.
Delete branch "feat/docs-gaps-top4"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Summary
Four small, atomic commits that close the four highest-priority documentation
gaps identified after the
feat/pr-get-and-createPR landed.chore(tooling): make bun test robust under sandbox limits—package.jsontest script now setsBUN_MAX_WORKERS=1, dodging Bun'sFailed to start HTTP Client thread: Resource temporarily unavailablepanic on CPU/thread-constrained sandboxes. AGENTS.md Gotchas records the
rationale and points CI-locals at
bun testdirect for permissiveenvironments.
docs(architecture): document real cwd inference with worktree support—Replaces the stale "this module is a stub" paragraph in
docs/ARCHITECTURE.mdwith an accurate description of thegit -C <cwd> remote get-url originimplementation, worktree handling(
.gitas a file withgitdir:vs. as a directory), SSH→https://<host>URL normalization, and the three-layer test strategy (pure URL parser,
mocked detection, end-to-end real-git fixture).
docs(agents): add branch naming convention— Branches followfeat/<scope>, matching Conventional Commits scope. Same convention forbranches and commits reduces cognitive load.
docs(prd): record resolved write-guard decision— Adds a "Resolveddecisions" section to
docs/PRD.mdcapturing why theconfirm: true/confirm_titleper-call guard and the process-level write flagwere removed in the previous PR.
Scope
Doku-only. No code or interface changes. CLI / MCP / OpenCode behavior is
identical to
main.Verify
bun run lint— greenbun run typecheck— greenbun run test— 122/122 green (now without manualBUN_MAX_WORKERS)bun run build— green (untouched by this PR)Out of scope
Lower-priority doc gaps noted in the inventory but not addressed here:
docs/GLOSSARY.mdstaleness check, ADR for the cwd-inference normalization,comment in
src/shared/cwd.tsexplaining the worktree mechanism. These remaincandidates for a future
feat/docs-gaps-restPR.