agentsclimarketplace

Ticket test design

Skill cat2000/domain-knowledge-ops/.cursor/skills/ticket-test-design

Design release-confidence test cases for one Jira story: must/should/later cut, AC traceability, ISTQB technique selection without checklist theater. Triggers: @ticket-test-design, Jira key, pasted AC, brief, mode=analysis, focus=security, or after @requirement-risk / @ticket-splitter on the same key. Reads domain briefs as evidence only; does not write curated.From its SKILL.md

Install
npx -y skills add cat2000/domain-knowledge-ops --skill ticket-test-design

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

7.4 KB, ~1.8k tokens by cl100k_base, as published. Nobody here has run it

ticket-test-design (single-ticket test spec)

First principles

Artifact: a ship/Done proof pack a tester can start running in ~60s — not an evidence memo.

Quality: every (given) AC has a decidable must (observable + oracle + seed) or Must-deferred / blocked readiness; must = contract instance, not partition exhaustion; same-AC residual disposed closed; lowest stable level that entails the AC; scan needed ⇒ closed disposition; should spend residual before chrome; honest automate. Presentation: short Summary, executable GWT, Design at end, chat = draft. Never invent product facts or framework code. Not a release-regression / metrics / NFR platform.

@requirement-risk@ticket-splitterThis skill
JobSprint decision inputINVEST backlog slicesTest evidence for Done/ship
Cut axisMUST/SHOULD/OPTIONAL severityItem boundariesmust / should / later anchored to given AC
GateStructure + countsTestable done_whenCounts + given-AC coverage + contract/pack + automate

Corollaries

  1. No human confirm loop by default (mode=analysis is the exception).
  2. Scripts check countable invariants (given-AC→must). They do not judge should priority — agent must apply should spend order + anti-padding from the rule.
  3. Substance: ../../rules/ticket_test_design.md. Readability: references/presentation.md.

When to use / not

UseDo not use → instead
Design cases for one ticket before/during QAReadiness only → @requirement-risk
Trace cases to AC / done_whenSplit backlog → @ticket-splitter
Cut infinite design space for releaseWrite Playwright/API automation → dedicated automation skill
Offline DEMO-* rehearsalRefresh briefs → @generate-knowledge-from-wiki

How to invoke

@ticket-test-design DEMO-1 team=demo
@ticket-test-design PROJ-123
@ticket-test-design team=<key> PROJ-123 brief
@ticket-test-design PROJ-123 mode=analysis
@ticket-test-design PROJ-123 focus=security

Or paste requirement / AC text.

ParamValuesDefault
teamkey/alias in team-roots.jsonattribution / Agile Team
briefshort-mode words in messagefull spec
modeanalysisgenerate cases
focussecuritybaseline scan

Offline: ../_shared/offline-demo.md. Team→root: ../_shared/team-root.md.

Workflow

Jira key or text?
├─ DEMO-* / offline → fixtures; no network
├─ real key → issue + attachments (MCP or fetch_jira_attachments.py)
├─ Resolve team/root → S7 locale brief as evidence (note if only S6/draft)
├─ Build acceptance (jira → split done_when → proposed)
├─ Pick primary technique (+ secondary if needed) — see technique-selection
├─ For each given AC: decidable must + closed residual + level
├─ mode=analysis? → deliver scope/design/acceptance → STOP
├─ Draft → .jira_attachments/<KEY>/test_design_draft.md (Environment: build + seed)
└─ validate → fix → deliver in chat **same Must GWT as the draft**

Responsibility

LayerAgentScript
EvidenceBriefs, dual-root, risk refsFetch issue/attachments
GenerateAC, cases, charters; decidability, level, should spend, scan disposalNo prose authorship
GateFix reported failures onlyvalidate_ticket_test_design.py (incl. given-AC coverage)
Soft should priorityAgent (rule spend order + anti-padding) + human readerNo

Gate rules

  1. Pass validator before claiming done (full and brief).
  2. Summary counts must match case headings (must / should / later).
  3. Summary has Contract readiness + Pack note (no legacy single Readiness line).
  4. Every (given) AC appears in some must proves or under Must-deferred (blocked contract readiness).
  5. Every must case has honest proves:; multi-AC ⇒ (AC-n) in then; must/should have automate:.
  6. should uses proves or supplements:; later = charter/idea only; never write curated/.

Agent checklist

  1. Load evidence per rule + ../../contracts/jira-issue-domain-knowledge-context.md.
  2. List all (given) AC; plan one decidable must each (oracle + seed); list partition sets and field modes; plan closed residual disposition.
  3. Set Contract readiness from given-AC must coverage and decidability; put should/weak oracles in Pack note.
  4. Choose level (api/logic/ui) per lowest stable interface; don’t default data ACs to UI-only.
  5. Should spend order: same-AC residual (closed) → weak-oracle hardening → security/resilience supplements → chrome last.
  6. If scan security|resilience: needed → closed disposition (should/later/risk).
  7. Write Must cores; Design after Later; Environment with device/build + seed: when shared; Summary = decision board.
  8. Honest automate; draft → validate → paste full indented Must in chat.
  9. Optional handoff: @requirement-risk / @ticket-splitter / automation skill — not codegen here.

Done when

  • Validator exit 0 (or explicitly marked failed)
  • Summary decision board readable in ≤60s
  • Every given AC covered by decidable must or Must-deferred / blocked-by-evidence
  • Same-AC residuals disposed closed
  • Scan needed disposed or truly out_of_scope
  • Environment names how to run (build + seed if needed); chat = draft
  • Honest automate + sensible level; no stretch proves; no full-coverage claim

Forbidden

  • MECE five-layer dumps / emoji Golden / table-compressed Must in chat
  • Summary evidence essays; Design before Must
  • Soft then without oracle_confidence: weak
  • Open residual shoulds; chrome should before same-AC residual
  • Undecidable must + contract-ready; needed scan with no disposition
  • Quantity quotas; inventing AC/steps; framework codegen
  • Expanding into release packs / metrics / default NFR-chaos platforms
  • Skip gate; confirm ritual unless mode=analysis / ungated draft requested

References (progressive)

Examples

@ticket-test-design DEMO-1 team=demo
→ offline fixture → draft → validate → readable spec

@ticket-test-design PROJ-123 brief
→ summary + AC + must only → validate --brief

What ships with it: 4 files

15.5 KB alongside SKILL.md

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.