oat-review-receive

name	oat-review-receive
version	1.4.0
description	Use when processing review findings outside project context. Converts local review artifacts into actionable task lists.
disable-model-invocation	true
user-invocable	true
allowed-tools	Read, Write, Bash, Glob, Grep, AskUserQuestion

Review Receive (Ad-hoc Local)

Process local review artifacts into a normalized findings register and generate actionable standalone tasks.

Prerequisites

• A review artifact exists locally (.md).
• User wants ad-hoc triage without mutating OAT project lifecycle artifacts.

Mode Assertion

OAT MODE: Review Receive

Purpose: Parse local review findings, classify severity consistently, triage disposition with user input, and generate a task list.

BLOCKED Activities:

• No implementation/code changes.
• No plan.md, state.md, or implementation.md lifecycle mutations.
• No silent dismissal/defer decisions.

ALLOWED Activities:

• Locating review artifacts.
• Parsing markdown findings into normalized records.
• Presenting findings counts and summaries.
• Interactive triage (convert, defer, dismiss).
• Writing standalone task-list output.

Self-Correction Protocol: If you catch yourself:

• Editing project lifecycle docs in ad-hoc mode -> STOP and revert to task-list output only.
• Triaging without presenting a findings overview first -> STOP and show overview before disposition prompts.
• Skipping rationale when proposing deferral at any severity (including Minor) -> STOP and collect explicit rationale.
• Defaulting a Minor finding to defer instead of convert -> STOP; the Minor default is convert (fix inline), and defer requires concrete rationale.

Recovery:

1. Re-locate review artifact from the path provided or last known location.
2. Re-parse findings from the artifact (idempotent — no state mutation in ad-hoc mode).
3. Resume triage from the first un-dispositioned finding.

Progress Indicators (User-Facing)

Print this banner once at start:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ OAT ▸ REVIEW RECEIVE ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Use these step indicators:

• [1/4] Locating review artifact...
• [2/4] Parsing findings...
• [3/4] Triaging findings...
• [4/4] Generating task list...

Findings Model

Normalize every finding to this shape:

finding:
  id: "C1" | "I1" | "M1" | "m1"
  severity: critical | important | medium | minor
  title: string
  file: string | null
  line: number | null
  body: string
  fix_guidance: string | null
  source: local_artifact
  source_ref: string

Severity conventions:

• critical: Missing P0 requirements, security vulnerabilities, broken behavior.
• important: Missing P1 requirements, major error-handling or maintainability gaps.
• medium: P2 issues with meaningful impact.
• minor: Low-impact polish/documentation/style issues.

ID conventions:

• Critical: C1, C2, ...
• Important: I1, I2, ...
• Medium: M1, M2, ...
• Minor: m1, m2, ...

Process

Step 1: Locate Active Review Artifact

Artifact source priority:

1. Explicit path from $ARGUMENTS
2. Most recent top-level file under .oat/repo/reviews/
3. Most recent top-level file under .oat/projects/local/orphan-reviews/

Discovery command example:

ls -t .oat/repo/reviews/*.md .oat/projects/local/orphan-reviews/*.md 2>/dev/null | head -20

Selection rules:

• Ignore .oat/repo/reviews/archived/ and .oat/projects/local/orphan-reviews/archived/.
• Treat archived review artifacts as history; do not re-triage them by default.
• If an explicit path points into an archived/ directory, stop and ask for an active review artifact instead.

If multiple candidates exist, auto-select the most recent by oat_generated_at frontmatter date (not filesystem mtime, which is unreliable across branches). Inform the user which artifact was selected and list any others for awareness. Do not prompt for selection.

Validation:

• File exists and is readable.
• Extension is .md (or confirm nonstandard markdown extension).
• Content is non-empty.

Step 2: Parse Findings

Parse by severity sections/headings using case-insensitive matching:

• Critical
• Important
• Medium
• Minor

Compatibility rule:

• If artifact uses a 3-tier model (no Medium section), treat Medium as zero findings.

Extraction guidance per finding item:

• Derive title from first concise clause/line.
• Extract file + line if present in common patterns (path:line, fenced diff context, inline references).
• Populate body with the finding detail.
• Populate fix_guidance when explicit fix direction exists.
• Set source: local_artifact.
• Set source_ref to the artifact path.

Step 3: Present Findings Overview

Before asking for dispositions, print:

• Total counts per severity.
• A compact finding register grouped by severity, each showing:
  • id
  • title
  • file:line (or -)

Example summary:

Critical: 1
Important: 2
Medium: 1
Minor: 3

If there are zero findings across all severities, output a clean result and stop.

Step 4: Interactive Triage

For each finding, ask for disposition:

• convert -> create standalone task entry
• defer -> keep out of current task list, record reason
• dismiss -> close without task, record reason

Default suggestions:

• Critical -> convert
• Important -> convert
• Medium -> convert (propose defer only with concrete rationale)
• Minor -> convert (propose defer only with concrete rationale)

Rules:

• Require explicit rationale for defer or dismiss at any severity, including Minor. Small findings are cheap to fix inline, so deferring a Minor into a backlog item must be justified just like any other deferral.
• Do not recommend defer for a Minor finding solely because it does not impact current functionality — fixing inline is usually cheaper than tracking it. Reserve defer for the genuine cases (low-probability cleanup, blocked dependency, duplicated elsewhere, explicitly out of scope, or disproportionate churn now).
• Do not silently skip findings.

Step 5: Generate Task List Output

Generate standalone markdown tasks (no plan task IDs):

- [ ] [critical] Fix auth bypass in token validator (`src/auth/token.ts:91`) - Enforce issuer/audience validation.

Output modes:

• Inline (default)
• File output path (if user requests), for example:
  • .oat/projects/local/orphan-reviews/review-tasks-YYYY-MM-DD.md

Also output deferred and dismissed findings with reasons.

Step 6: Archive the Consumed Review Artifact

After triage is complete, move the processed review artifact into a sibling archived/ directory and report the archived path:

• .oat/repo/reviews/{file}.md -> .oat/repo/reviews/archived/{file}.md
• .oat/projects/local/orphan-reviews/{file}.md -> .oat/projects/local/orphan-reviews/archived/{file}.md

Rules:

• Create the archived/ directory if needed.
• If the archive target already exists, append a timestamp suffix before moving so prior history is not overwritten.
• Keep the archived path in the final output contract.

Output Contract

At completion, report:

• Artifact path used
• Archived artifact path
• Counts by severity
• Number converted/deferred/dismissed
• Task list location (inline or file path)

Success Criteria

• Local artifact resolved and validated.
• Findings parsed into 4-tier normalized structure.
• Findings overview displayed before triage.
• Every finding dispositioned with rationale where needed.
• Standalone task list generated in requested output mode.
• Consumed review artifact archived in the matching archived/ directory.
• Skill remains within content budget (<=500 lines).