Skip to content

checkpoint command

Command

Periodic multi-reviewer sweep of the whole codebase — surfaces a triaged checkpoint report.

Invoke it
Claude Code/ca:checkpoint
Codex$ca-checkpoint
Pi/ca-checkpoint

Availability is read from each shipped host catalog. Preview-only commands are marked on the page.

A periodic whole-repository review. The caller supplies the reviewer unit list to dispatching-parallel-agents; the read-only results pass through finding-triage and verdict-aggregator. Only after that verdict returns does the caller separately dispatch checkpoint-aggregator to persist every finding and incomplete-unit result.

The report uses .codearbiter/checkpoints/YYYY-MM-DD.md, with -2, -3 suffixes when necessary. Existing reports are never overwritten. The command then writes the nonblank, noncomment override count to .codearbiter/last-checkpoint, re-zeroing the overrides-since-checkpoint display. That integer is neither a timestamp nor a commit-gate receipt. A periodic report provides no promotion sign-off and does not satisfy typed task acceptance.

The caller captures the reviewed commit, UTC review-start time and whether the worktree was dirty before dispatch. The report retains these as source_commit, recorded_at and worktree_dirty; missing observations remain unknown. Audit windows use these report anchors, not the counter.

Checkpoints in context explains the complete handoff and how to inspect missing review evidence before routing follow-up work.

Published releases from 0.7.5 include each packaged reviewer resource charter for Codex host-provided agent threads. Exact static-package and route-closure checks gate release. A bounded 0.9.4 receipt proves one exact installed-charter review. The inline fallback applies only where the canonical workflow explicitly permits it and isolation is not mandatory. See Claude Code + Codex → Intentional host differences.

/ca:checkpoint

Takes no arguments — it sweeps the whole codebase every time.

Illustrative output, not captured execution:
> /ca:checkpoint
Read-only reviewer batch -> triage -> verdict.
Separate checkpoint writer retains the complete verdict.
Selected report: .codearbiter/checkpoints/2026-09-24-2.md
last-checkpoint: 12 (override-count baseline)
No code repair or delivery sign-off performed.

The routine whole-codebase sweep — not the current diff (/ca:review), not the rare deep audit (/ca:tribunal), and not ADR health alone (/ca:adr-status).

GateWhenEffect
reviewer fleet, funneledevery invocationsecurity-reviewer, auth-crypto-reviewer, dependency-reviewer, migration-reviewer, coverage-auditor, and architecture-drift-reviewer run read-only over the whole codebase, then return a finding-triage and verdict-aggregator verdict before the caller separately dispatches the checkpoint writer

Find this entry in the workflow atlas. Compare its related teaching maps, host spelling and context inputs. The exact procedure below remains the owner.

Source — plugins/ca/commands/checkpoint.md (v3.7.0)
---
description: Periodic multi-reviewer sweep of the whole codebase — surfaces a triaged checkpoint report.
argument-hint: (none)
---
# /ca:checkpoint — codebase sweep
Periodic sweep of the entire codebase with the reviewer fleet, funneled to a single dated report. Surfaces findings — not a promotion gate, enforces no sign-off.
## Flow
1. Build the unit list — the same fleet as `/ca:review`, scoped to the whole codebase against the
`${CLAUDE_PROJECT_DIR}/.codearbiter/` docs:
| Reviewer | Reads |
|---|---|
| `security-reviewer` | `security-controls.md`; reviews security posture |
| `auth-crypto-reviewer` | `security-controls.md`; authn/crypto/key/secret paths |
| `dependency-reviewer` | dependency manifests; license + supply-chain posture |
| `migration-reviewer` | migration history; safety + classification |
| `coverage-auditor` | test coverage vs. obligations across the tree |
| `architecture-drift-reviewer` | `decisions/`; drift between code and accepted ADRs |
2. Route to `dispatching-parallel-agents` (`${CLAUDE_PLUGIN_ROOT}/skills/dispatching-parallel-agents/SKILL.md`) with that unit list (read-only batch). It dedupes, then
funnels through `finding-triage` → `verdict-aggregator` and returns the single read-only verdict.
3. After the verdict returns, separately dispatch `checkpoint-aggregator` with that verdict and the reviewed commit from `git rev-parse HEAD`, the UTC review-start time, and whether the reviewed worktree was dirty. Retain these observed values from before the sweep. It writes the dated report to
`${CLAUDE_PROJECT_DIR}/.codearbiter/checkpoints/YYYY-MM-DD.md`: findings by severity with
file:line, and out-of-scope items marked inline `[NEEDS-TRIAGE]`.
4. Write the current override **count** to `${CLAUDE_PROJECT_DIR}/.codearbiter/last-checkpoint` — the
integer baseline the statusline subtracts for its overrides-since-checkpoint counter. The value is
the number of non-comment, non-blank lines in `overrides.log` at this moment (`0` if the log is
absent). This re-zeros the `over:N` segment until the next `/ca:override`. Write a bare integer,
not a timestamp — the statusline treats any value above the current total as stale and falls back
to showing every override.
5. Report the checkpoint path.
## Hard gate
Read-only except writing the checkpoint doc and `last-checkpoint` — MUST NOT modify code. MUST NOT
consume raw reviewer output — only the `finding-triage` → `verdict-aggregator` verdict. Checkpoint
persistence MUST remain the separate `checkpoint-aggregator` step and MUST NOT run for `/ca:review`
or another generic parallel batch. MUST NOT resolve a `[CONFIRM-NN]` surfaced during the sweep by guessing. The report surfaces findings; it does
not block or sign off anything.
## When NOT to use
- Reviewing just the current diff → `/ca:review`.
- A pre-implementation threat model → `/ca:threat-model`.
- ADR health only → `/ca:adr-status`.
- A whole-codebase deep audit → `/ca:tribunal` (checkpoint is the lean periodic sweep; tribunal its rare deep counterpart).

View in repo