agentleFS
Codagent-AICodagent-AI/agent-skills/skills/test-plan/SKILL.md

test-plan

Skill31 starsChanged 6 months agoMITView on GitHub

Collaboratively creates a risk-based test plan for a defined software change using the automated test pyramid plus the envelope an exploratory acceptance pass runs under and exceptional human-only testing; use after specifications and design are complete, when asked to plan testing, define integration or end-to-end coverage, set testing authorizations, or create test-plan.md before implementation planning.

We split this skill into the rules it gives an agent

7 rules. 0 are common, 4 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 14 edited versions we read keep these.

  • 4

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Create <change-dir>/test-plan.md after the proposal, specifications, and design are complete.

    line 7
  2. Very specific:

    Do not write the plan until the user approves the proposed coverage.

    line 12
  3. Distinctive:

    Read the definition artifacts, relevant repository instructions, current test structure, and affected public surfaces.

    line 18
  4. Distinctive:

    Prefer controlled real dependencies or contract tests when they are more faithful than mocks.

    line 26
  5. Distinctive:

    Avoid fixed ratios, test-count quotas, duplicate assertions across layers, and E2E coverage for behavior a cheaper layer proves adequately.

    line 30
  6. Distinctive:

    Do not enumerate acceptance test cases.

    line 42
  7. Very specific:

    Default human-only testing to None.

    line 59

“Common” is measured: kept by at least 6 in 10 of the 14 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Test Plan

Create <change-dir>/test-plan.md after the proposal, specifications, and design are complete. The plan records important automated integration and end-to-end obligations, the envelope the later exploratory acceptance pass runs under, and exceptional human-only checks. Specifications and implementation-time decisions remain the source of unit-test requirements.

Do not write the plan until the user approves the proposed coverage. Use codagent:ask-questions for consequential choices involving environments, external effects, cost, credentials, fidelity, substitutes, or genuinely human-only judgment.

Plan coverage by risk

Read the definition artifacts, relevant repository instructions, current test structure, and affected public surfaces. Trace critical journeys and boundaries, then choose the lowest test layer that can reliably detect each important failure:

  • Unit tests: the broad base for isolated logic, validation, transformations, decisions, and edge cases. Do not inventory these in the test plan.
  • Integration tests (INT-*): important boundaries where real components must work together, such as adapters, databases, filesystems, subprocesses, APIs, queues, or configuration-to-runtime wiring. Prefer controlled real dependencies or contract tests when they are more faithful than mocks.
  • Automated end-to-end tests (E2E-*): a small set of critical journeys through a public entry point with realistic isolated setup and stable observable assertions.

Avoid fixed ratios, test-count quotas, duplicate assertions across layers, and E2E coverage for behavior a cheaper layer proves adequately. It is valid to record that no new integration or E2E obligation is warranted and explain why.

Plan automated tests for code only, never for prompts. Do not plan a test that checks whether a prompt, such as a SKILL.md, contains certain text.

For each automated obligation, capture what it covers, the boundary or journey exercised, setup, stable assertions, constraints, and where it runs.

Define the testing envelope

Do not enumerate acceptance test cases. Acceptance is an exploratory pass: it exercises each user-visible requirement or scenario the approved specs add or modify at least once, derived from the specs themselves, and sizes further exploration to what actually changed. A pre-written flow list only tells it to re-walk the happy path. The plan's job is to say what that pass is allowed to do, not what it must check.

Record the envelope:

  • environments, sandboxes, and test accounts available;
  • credentials and secrets that exist, and where;
  • external effects that are authorized, their cost, and required cleanup;
  • anything the pass must not touch, such as production data or irreversible operations;
  • substitutes permitted when a real dependency is unavailable.

Note known risk areas, prior defect clusters, and anything the design treats as an accepted limitation. These inform exploration without prescribing it.

Default human-only testing to None. Add an HT-* only for judgment or authority unavailable to an agent with the product and tools—for example subjective preference, legal approval, physical perception, unavailable personal credentials, or an irreversible user-authorized act. Visual and interactive UI testing belongs to the exploratory pass. For each HT-*, state why an agent cannot perform it, what prior testing must establish, concise user instructions, and the decision or observation required.

Approve and write

Present the proposed obligations, the envelope, meaningful omissions, and consequential testing choices. After user approval, write the plan and a coverage map containing only requirements or journeys with an additional INT-*, E2E-*, or HT-* obligation. Do not create tasks, write tests, execute tests, or invoke another lifecycle phase.

Artifact template

## Coverage Strategy

Specifications remain the source of unit-test requirements. This plan records only additional
integration and end-to-end obligations, the acceptance testing envelope, and exceptional human-only
obligations.

## Integration Tests

### INT-001: <boundary behavior>
- Covers: <requirements or journey>
- Boundary: <real components and interaction>
- Setup: <controlled dependencies and data>
- Action: <operation>
- Assertions: <stable observable results>
- Execution: <test location or CI command/phase>

## End-to-End Tests

### E2E-001: <critical journey>
- Covers: <requirements or journey>
- Surface: <public entry point>
- Setup: <realistic isolated environment>
- Journey: <user/client actions>
- Assertions: <stable observable results>
- Execution: <test location or CI command/phase>

## Acceptance Testing Envelope

- Environments and sandboxes: <available targets>
- Credentials and secrets: <what exists and where>
- Authorized effects: <external effects, cost, and required cleanup>
- Off limits: <what must not be touched>
- Permitted substitutes: <substitute and condition, or None>
- Known risk areas: <prior defect clusters and accepted limitations>

## Human-Only Testing

None.

<!-- When needed:
### HT-001: <human judgment or authority>
- Reason: <why an agent cannot perform it>
- Prerequisites: <prior automated and acceptance evidence>
- Instructions: <concise user actions>
- Required decision or observation: <result needed>
-->

## Coverage Map

| Requirement or journey | INT | E2E | HT |
| --- | --- | --- | --- |
| <item> | <IDs or —> | <IDs or —> | <IDs or —> |

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.