Review To Test Planner

Plan follow-up test work from review findings. This skill sits between review artifacts and

test-generator

: it translates findings into bounded test briefs, but it does not write tests. In short, it is the bridge between review artifacts and

test-generator

.

Use this skill when CodeBuddy or an operator has findings from repo review, security review, architecture review, dependency review, docs review, synthesis, or third-party review and wants a source-grounded plan for what should be tested next.

Quick Start

1. Confirm the bounded review target:
   • review packet
   • specialist artifact directory
   • synthesis artifact
   • single review file
2. Prefer CodeBuddy packet artifacts when available:

   • evidence_index.json

   • repo_inventory.json

   • run_manifest.json

   • specialist review artifacts
3. Run the deterministic planner when local access is available:

   • scripts/plan_review_tests.py <review-root> --out <artifact-root>

4. Inspect the generated plan and tighten any findings that need human judgment.
5. Hand only safe, concrete tasks to

   test-generator

   . Stop before writing tests, fixtures, issues, PRs, or customer-repo changes.

Focus

Prioritize:

• behavior under test, not generic coverage advice
• the smallest meaningful test target for each finding
• suggested test location and framework based on source path evidence
• fixture needs and negative cases
• expected assertions and validation commands
• whether the finding is safe for automated test generation
• explicit

  test-generator

  handoff blocks for safe tasks

Classify each task with one generation status:

• generated

  : a complete handoff brief already exists and can be passed to

  test-generator

  without additional planning.

• recommended

  : enough evidence exists to recommend a bounded test-generation task, but a human/operator should still decide whether to run it.

• deferred

  : more evidence, implementation context, or product decision is needed before test generation.

• unsafe

  : automated test generation should not proceed because the task would risk secrets, production systems, destructive behavior, privacy, or unbounded repo mutation.

Defer primary ownership of these areas:

• writing tests or fixtures:

  test-generator

• identifying original findings: review specialist skills
• fixing production code: implementation issue workflow
• creating issues from findings: finding-to-issue planner
• final synthesis/report writing: synthesis or report writer skills

Required Inputs

At minimum, gather:

• repo_root

• one concrete target:

  • target.review_packet_path

  • target.review_artifact_path

  • target.specialist_artifacts

  • target.findings_file

Useful additional inputs:

• artifact_root

• diff_base

• changed_paths

• test_framework_hint

• validation_mode

• generation_policy

• allowed_test_roots

• blocked_test_roots

If there is no bounded review artifact or finding source, stop and report

blocked

.

Workflow

1. Establish Scope

Record:

• reviewed artifact paths
• finding sources considered
• evidence packet consulted
• test framework hints
• blocked paths or unsafe domains

Do not widen a single review artifact into a whole-repo test strategy unless the input is explicitly a whole-repo review packet.

2. Map Findings To Behavior

For each finding, identify:

• priority and title
• affected file or subsystem
• behavior under test
• risk or regression scenario
• source evidence
• likely test framework and test location
• fixture/setup needs
• core assertions
• validation command

If the behavior or target file cannot be named concretely, mark the task

deferred

.

3. Classify Generation Safety

Mark a task

unsafe

when test generation would require:

• real credentials, secrets, or production accounts
• destructive filesystem, database, network, billing, or deployment actions
• unbounded customer-repo mutation
• broad refactors rather than focused tests
• guessing behavior not supported by review evidence

Only safe tasks should include a ready

test-generator

handoff.

4. Emit Handoffs

For

generated

and

recommended

tasks, include a structured handoff that can be reviewed before invoking

test-generator

.

The handoff should specify:

• skill_input_schema: test_generator.v1

• mode
• repo root
• target file/path/worktree/diff
• target behavior
• acceptance surface
• test depth
• fixture policy
• validation mode

• stop_after_generation: true

Do not invoke

test-generator

unless the operator explicitly asks for the follow-on execution.

Output Expectations

Default output should include:

• findings-to-test map
• generation-status summary
• test task briefs
• fixture and assertion map
• validation command plan
• safe

  test-generator

  handoffs
• deferred and unsafe tasks
• validation performed or not run
• residual test-planning risk

Use

references/output-contract.md

and the shared suite contract in

adl/tools/skills/docs/MULTI_AGENT_REPO_REVIEW_SKILL_SUITE.md

.

Stop Boundary

Stop after producing the test planning artifact.

Do not:

• write tests, fixtures, snapshots, or production code
• mutate customer repositories
• create issues or PRs
• call

  test-generator

  automatically
• claim tests are generated when only a plan exists
• replace review specialists, synthesis, finding-to-issue planning, or implementation workflow

CodeBuddy Integration Notes

This skill consumes CodeBuddy review packets and specialist artifacts. It produces a review-to-test plan that can feed

test-generator

, synthesis, product reports, or milestone follow-up triage.

Deferred automation:

• richer parsing for third-party PDF review extracts
• repository-specific framework discovery from package manifests
• direct confidence scoring from executed coverage data
• optional batch handoff execution through a separate conductor-approved flow