create-manual-test-instructions
Skill85 starsChanged 21 days agoMITView on GitHub
Turn a ticket, requirements document or code change (branch, PR, commit range, diff) into a concise QA manual-test file a non-author can follow. Invoke manually only.
In this repository31 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- agentify-projectSkill
- attach-to-ticketSkill
- check-ticket-implementationSkill
- compact-docs-writerSkill
- compact-skill-creatorSkill
- context-checkupSkill
- create-implementation-planSkill
- execute-plan-tasksSkill
- explain-in-simple-languageSkill
- extract-pr-review-rulesSkill
- fetch-pr-reviewSkill
- fetch-ticketSkill
- fresh-eyes-reviewSkill
- grill-meSkill
- handoverSkill
- harden-artifactSkill
- maintainer-reviewSkill
- memory-doctorSkill
- prepare-promptSkill
- refine-pr-reviewSkill
- refine-ticketSkill
- review-code-assistantSkill
- review-ticketSkill
- run-nx-checksSkill
- self-improveSkill
- self-reviewSkill
- split-plan-tasksSkill
- use-conversational-languageSkill
- verify-understandingSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 70
Before writing, actually invoke explain-in-simple-language and follow it — reciting its rules from memory does not count.
- Distinctive: line 74
Do not modify any source files.
- Distinctive: line 76
When done, state the manual-test file's project-relative path and, for a branch or PR, the target and the source state (tip or working …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Create manual test instructions
Resolve the input
- Ticket or requirements document (file or pasted text) — the only source: the checks follow it, and no code change is looked up.
- Code change — a branch, PR URL, commit range, pasted diff or patch file, or nothing (the
current branch). A commit range is diffed between its own ends. A branch or PR:
git fetchthe remotes, then diff from the merge base<base>with the target — a PR's base from its metadata, else the default branch of theupstreamremote when one exists, else oforigin, viagit ls-remote --symref <remote> HEAD; that failing, the sole existing candidate amongmain,master,develop/development, preferringupstream's remote-tracking ref, thenorigin's, then the local branch.git diff <base> <source>for a tip; a checked-out source is reviewed as its working tree:git diff <base>plus the untracked files (git ls-files --others --exclude-standard), read whole. Read the commit subjects and the PR title and description too. Then resolve the task's planning home — its<id>-<slug>/directory (no ticket → no id) or flat<id>-<slug>.*files in the project's planning directory, id and slug from the branch name or PR — and read the ticket and requirements there when present for the intent, the "before" and the edge cases; the checks still follow the code, a Reproduction part's steps aside (What to verify). No merge base →git fetch --unshallowand retry. Running forked, you can't ask mid-run: an ambiguous target, a failed fetch, still no merge base or an empty diff → write nothing and say in the final message what to pass on the next run; never guess a target.
Where needed for accurate navigation, also read the code the input cites and the design frames or screenshots it links when they are on disk.
Write the file
Next to an input ticket or requirements file, naming it by replacing .REQUIREMENTS or .TICKET
with .MANUAL-TEST (FOO.REQUIREMENTS.md → FOO.MANUAL-TEST.md); neither convention → append
.MANUAL-TEST before .md. A code change with a planning home → <id>-<slug>.MANUAL-TEST.md
beside the task's other artifacts. Otherwise (pasted text, a diff or patch file, a branch without
artifacts) follow the project's/user's convention for where plans live (default: .agents/plans/)
and write <id>-<slug>/<id>-<slug>.MANUAL-TEST.md there, id and slug from the branch name or PR,
else a kebab-case slug from the content. An existing manual-test file of that name is replaced;
the final message says so. Flag any guessed location or slug in the final message for the user to
verify.
The file must be followable by someone unfamiliar with the ticket — short but complete. Four parts:
- What changed — 1–2 sentences: the feature and the user-visible difference.
- How to get there — concrete navigation to the affected area (entry point, page/screen name, any prerequisite state).
- Before vs after — how the area behaved before, how it should behave now.
- What to verify — checklist of behaviors to confirm, including edge cases (empty values, multiple items, boundaries) where relevant. A requirements document's Reproduction part, in the input or read for a code change, gives the first check, a regression check: its steps, passed when the expected behaviour it records shows. A ticket's own reproduction field isn't that part. A behaviour resting on an Open question of the input goes under a closing Not settled yet line, never as a plain check: on a non-blocking question, with its tentative answer; on a blocking one, listed as undecided, naming who decides when the input names them, with no expected value — never invented. Each point where a code change differs from its ticket — built but never asked for, or asked for but not built — goes under a closing Differs from the ticket line, with nothing invented for the missing side.
A change with nothing a person can test by hand (a refactor, a dependency bump, a backend change with no visible effect) still gets the file: part 1 says why, the other parts stay empty.
Before writing, actually invoke explain-in-simple-language and follow it — reciting its rules from memory does not count. Not installed: plain words, one idea per sentence, and nothing named only by a stand-in the reader hasn't seen.
Do not modify any source files; the only file you write is the manual-test document.
When done, state the manual-test file's project-relative path and, for a branch or PR, the target and the source state (tip or working tree) you diffed.
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.

