126 lines
6.7 KiB
TypeScript
126 lines
6.7 KiB
TypeScript
import type { WorkflowIrNode } from "./workflow-ir-types.js";
|
|
|
|
/*
|
|
FNXC:CodeReviewStep 2026-06-25-15:00:
|
|
Code Review is a DEFAULT-ON but TOGGLEABLE step in the built-in coding and
|
|
stepwise-coding workflows: an `optional-group` container node with `defaultOn: true`.
|
|
It is part of the existing flows and runs for every coding task by default (the
|
|
default-on resolver seeds `code-review` into a new task's enabledWorkflowSteps), yet an
|
|
operator can turn it off per task by removing `code-review` from enabledWorkflowSteps —
|
|
when disabled the group passes through byte-inert, restoring the exact prior flow.
|
|
|
|
The group sits on the pre-merge success path (execute → [browser-verification optional]
|
|
→ code-review → review). The group node id `code-review` is the STABLE per-task enable
|
|
key; the inner template node carries a DISTINCT id (`code-review-step`) because a template
|
|
node id may not collide with the group/top-level node id (U1 validation).
|
|
|
|
The inner node is a blocking gate: a REVISE verdict must stop review/merge until
|
|
the executor remediates the finding or the task exhausts its revision budget.
|
|
|
|
FNXC:CodeReviewStep 2026-06-25-00:00:
|
|
U6 deleted the built-in step-template catalog; the inner node's literal
|
|
name/description/prompt/toolMode/gateMode are now inlined here directly. These built-ins
|
|
are the parity oracle; intentional behavior changes must update this file and the
|
|
built-in workflow tests together.
|
|
|
|
FNXC:CodeReviewStep 2026-06-29-10:42:
|
|
Code Review is not advisory. FN-7228 reached merge after Code Review returned REVISE
|
|
because the generic built-in was authored as advisory and the graph continued after the
|
|
remediation budget was exhausted. Keep browser verification advisory, but make Code
|
|
Review a gate so REVISE records a blocking failed workflow step and cannot advance to
|
|
review or merge.
|
|
|
|
FNXC:CodeReviewStep 2026-06-29-17:55:
|
|
Code Review REVISE is ordinary repair feedback, not a terminal task failure. FN-7242
|
|
exhausted the old built-in three-pass default and parked failed at
|
|
`code-review-remediation`; default built-in Code Review must keep recovering until it
|
|
passes unless a workflow author explicitly sets a numeric maxRevisions cap.
|
|
*/
|
|
|
|
/** Stable per-task enable key + group node id. */
|
|
export const CODE_REVIEW_GROUP_ID = "code-review";
|
|
|
|
/** Inner template node id — distinct from the group id (template-node-id collision rule, U1). */
|
|
export const CODE_REVIEW_STEP_NODE_ID = "code-review-step";
|
|
|
|
/** Display name (inlined from the former `code-review` catalog template). */
|
|
const CODE_REVIEW_NAME = "Code Review";
|
|
|
|
/** Short description (inlined from the former catalog template). */
|
|
const CODE_REVIEW_DESCRIPTION =
|
|
"Diff-review the task's changes for correctness bugs, regressions, and intent mismatches that tests miss";
|
|
|
|
/** Agent prompt (inlined verbatim from the former catalog template — parity oracle). */
|
|
const CODE_REVIEW_PROMPT = `You are a senior code reviewer. Review the task's diff for the correctness value automated tests do NOT catch.
|
|
|
|
## Step 1: Read the change
|
|
1. Read the full diff against the base branch: \`git diff <base>...HEAD\` (or \`git diff <base>\`). Determine the base from the task context / merge target.
|
|
2. Read the changed files in full where the diff is non-trivial, so you see the surrounding code paths the change touches — not just the hunks.
|
|
|
|
## Step 2: Review focus (the value tests miss)
|
|
1. **Correctness / logic bugs** — wrong conditions, inverted boolean/comparison logic, off-by-one, incorrect operator precedence, mishandled return values.
|
|
2. **Broken edge cases** — empty/undefined/null inputs, zero/duplicate/boundary values, concurrency and ordering assumptions.
|
|
3. **Intent vs implementation** — does the code actually do what the task/PROMPT.md describes? Flag silent scope drift or partial implementations.
|
|
4. **Regressions in touched code paths** — does the change break or weaken an existing behavior in the files it edits or their callers?
|
|
5. **Error handling** — swallowed errors, unhandled rejections/exceptions, missing validation at trust boundaries, misleading error messages.
|
|
6. **Contract / signature changes** — changed function/exported-type signatures, API request/response shapes, or serialization that breaks existing callers.
|
|
|
|
Be specific: cite \`file:line\` for every finding and explain the concrete failure it causes.
|
|
|
|
## Output Requirements
|
|
- Fast-bail: if the diff is trivial, generated, or out-of-scope for code review (e.g. pure docs/config/formatting with no logic), output {"verdict":"APPROVE","notes":"out of scope: code review"} immediately and stop.
|
|
- APPROVE: no correctness concerns; use empty or brief notes.
|
|
- APPROVE_WITH_NOTES: shippable, but include non-blocking advisories (with file:line) in notes.
|
|
- REVISE: a correctness bug, regression, or contract break requires changes; include file:line and the concrete failure plus remediation in notes.
|
|
- Final output: output exactly one trailing JSON object on the final line (no markdown fences, no surrounding prose):
|
|
{"verdict":"APPROVE|APPROVE_WITH_NOTES|REVISE","notes":"..."}`;
|
|
|
|
/**
|
|
* Build the `code-review` optional-group node placed on a workflow's pre-merge path.
|
|
* `defaultOn: true` makes it run by default while remaining togglable per task. `column`
|
|
* matches where the browser-verification group sits (in-progress) so the editor renders
|
|
* the group in the implementation column.
|
|
*
|
|
* A single `prompt` node whose config carries the inlined prompt +
|
|
* `toolMode: "readonly"` + blocking `gateMode: "gate"`.
|
|
*/
|
|
export function codeReviewOptionalGroupNode(
|
|
column: string,
|
|
options: { defaultOn?: boolean; maxRevisions?: number | "unbounded" } = {},
|
|
): WorkflowIrNode {
|
|
return {
|
|
id: CODE_REVIEW_GROUP_ID,
|
|
kind: "optional-group",
|
|
column,
|
|
config: {
|
|
name: CODE_REVIEW_NAME,
|
|
// Default-ON: runs for every coding task by default, but operators can toggle it
|
|
// off per task (remove `code-review` from enabledWorkflowSteps).
|
|
defaultOn: options.defaultOn ?? true,
|
|
reworkRegion: true,
|
|
maxReworkCycles: 3,
|
|
/*
|
|
* FNXC:WorkflowRemediationBudget 2026-06-29-17:55:
|
|
* Built-in Code Review should never terminal-fail merely because repair feedback repeated. Default to unbounded fix→review recovery while preserving workflow-authored numeric caps through `config.maxRevisions`.
|
|
*/
|
|
maxRevisions: options.maxRevisions ?? "unbounded",
|
|
template: {
|
|
nodes: [
|
|
{
|
|
id: CODE_REVIEW_STEP_NODE_ID,
|
|
kind: "prompt",
|
|
config: {
|
|
name: CODE_REVIEW_NAME,
|
|
description: CODE_REVIEW_DESCRIPTION,
|
|
prompt: CODE_REVIEW_PROMPT,
|
|
toolMode: "readonly",
|
|
gateMode: "gate",
|
|
},
|
|
},
|
|
],
|
|
edges: [],
|
|
},
|
|
},
|
|
};
|
|
}
|