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
- Very specific: line 7
Create <change-dir>/test-plan.md after the proposal, specifications, and design are complete.
- Very specific: line 12
Do not write the plan until the user approves the proposed coverage.
- Distinctive: line 18
Read the definition artifacts, relevant repository instructions, current test structure, and affected public surfaces.
- Distinctive: line 26
Prefer controlled real dependencies or contract tests when they are more faithful than mocks.
- Distinctive: line 30
Avoid fixed ratios, test-count quotas, duplicate assertions across layers, and E2E coverage for behavior a cheaper layer proves adequately.
- Distinctive: line 42
Do not enumerate acceptance test cases.
- Very specific: line 59
Default human-only testing to None.
“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.
Files
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 —> |Other versions17 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 17
- liferay/liferay-portal★ 2.3ksame name
- stackrox/stackrox★ 1.3ksame name
- MillenniumDawn/Millennium-Dawn★ 157same name
- mbwsims/claude-universe★ 71same name
- katalon-labs/true-skills★ 10same name
- yasunori0418/skills★ 9same name
- xmh-test/test-agent-skills★ 5same name
- DoKu67/claude-code-skils★ 4same name
- SUDARSHANCHAUDHARI/QACraft★ 2same name
- TzolkinB/skills★ 2same name
7 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
Used alongsideProjects with it also use reflect.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
Microsoft AzureSkillreflect
Learning capture system that extracts HIGH/MED/LOW confidence patterns from conversations to prevent repeating mistakes. Use after user corrections ("no", "wrong"), praise ("perfect", "exactly"), or when discovering edge cases. Complements .squad/agents/{agent}/history.md and .squad/decisions.md.
Also in 3 projects that use this
In this repository30 other files it gives its agents.
- AGENTS.mdAGENTS.md
- docs/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- releaseSkill
- releaseSkill
- ask-questionsSkill
- call-agentSkill
- designSkill
- finalize-prSkill
- fix-prSkill
- handoffSkill
- implement-changeSkill
- initSkill
- orchestrate-changeSkill
- plan-tasksSkill
- prepare-acceptanceSkill
- proposal-reviewSkill
- proposeSkill
- push-prSkill
- review-approachSkill
- review-assumptionsSkill
- review-specSkill
- review-tasksSkill
- session-reportSkill
- simple-planSkill
- specSkill
- task-complianceSkill
- test-flowsSkill
- verify-behaviorSkill
- wait-ciSkill
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
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.

