| name | code-review-all |
| description | Run a full code review of the current change — fan out every relevant review concern across independent subagents (the local review skills plus the ably-skills code-review-* plugin) and synthesise one consolidated report. Read-only; never edits code. |
| license | proprietary |
| metadata | {"team":"engineering","tags":"code-review, orchestration, fan-out, quality","mcp":false} |
| allowed-tools | Agent, Skill |
| argument-hint | [scope] (optional: a path, 'staged', 'repo', or a base ref; defaults to branch vs PR target/main) |
Code Review: all concerns
Run the full code review for the current change: fan out every relevant concern
across independent subagents — each running one focused review skill — then
synthesise their reports into a single consolidated review. This skill only
reviews; it never edits code.
Step 1 — resolve the base once
Resolve the base to review against, once, and reuse it for every subagent so
they all audit the same change:
- If
$ARGUMENTS carries a scope override (a path, staged, repo, or an
explicit base ref), use that verbatim.
- Otherwise resolve the current branch's open-PR target branch by invoking the
ably-skills:gh-resolve-target-branch skill; fall back to main if it reports
no PR.
Step 2 — fan out, one subagent per concern
Launch one general-purpose subagent per concern below, all in a single
message so they run in parallel. Give each subagent this prompt (substituting
the skill and the base from Step 1):
Run the <skill> skill (via the Skill tool) with scope argument <base> to
review the current change. Follow the skill in full and return only the
report it produces, ending in its single VERDICT: line. Do not edit anything.
Generic concerns — the shared ably-skills plugin skills:
| Concern | Skill |
|---|
| Back-compat | ably-skills:code-review-back-compat |
| Clean code | ably-skills:code-review-clean |
| Comments | ably-skills:code-review-comments |
| Correctness | ably-skills:code-review-correctness |
| Dead code | ably-skills:code-review-dead-code |
| DRY | ably-skills:code-review-dry |
| Error codes | ably-skills:code-review-error-codes |
| Logging | ably-skills:code-review-logging |
| Simplicity | ably-skills:code-review-simplicity |
| Tests | ably-skills:code-review-tests |
Project-specific concerns — the local skills in this repo:
| Concern | Skill |
|---|
| Demos | code-review-demos |
| Documentation | code-review-docs |
| Rules conformance | code-review-rules-conformance |
| Rules drift | code-review-rules-drift |
Skip a concern only when it plainly cannot apply to the change (e.g. no demos
touched). If you skip one, say so in the report rather than dropping it silently.
Step 3 — synthesise one report
Collect the subagents' reports and synthesise — do not just concatenate:
- Open with a summary table: one row per concern with its verdict
(PASS / CHANGES REQUESTED) and its blocker/major/minor/nit counts.
- List findings grouped by severity (all blockers first, then majors, then
minors, then nits), each tagged with its concern and
path:line. De-duplicate
findings several concerns raised about the same site, noting which concerns
flagged it.
- Close with the overall verdict:
VERDICT: CHANGES_REQUESTED if any concern
requested changes (any blocker or major across the board), otherwise
VERDICT: PASS.
If a subagent fails or returns nothing, note it in the summary table as
(no report) rather than dropping the concern silently. Apply no fixes — this
skill only reviews.