agentleFS
idaibinidaibin/skills/skills/dev-tests/SKILL.md

dev-tests

Skill6 starsChanged 20 days agoApache-2.0View on GitHub

Plan risk-based coverage and execute functional acceptance across backend, APIs, pages, UI, business logic and end-to-end journeys. Run existing scoped checks; route source fixes to their owner.

The file

Show its rules and text

We split this skill into the rules it gives an agent

13 rules. About 0 are common, 13 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈13

    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
  1. Distinctive:

    When asked to execute functional acceptance, run the reachable authorized cases.

    line 20
  2. Distinctive:

    Read current project commands and their effects before execution.

    line 50
  3. Distinctive:

    Read Playwright E2E only for coordinating generated browser-test acceptance.

    line 81
  4. Distinctive:

    Read coverage when deriving or reconciling the acceptance matrix.

    line 82
  5. Distinctive:

    Read execution before running checks or making regression claims.

    line 83
  6. Distinctive:

    Read functional workflow for functional campaigns and resumed case inventories.

    line 84
  7. Distinctive:

    Read API and logic for network, business status and persistence.

    line 85
  8. Distinctive:

    Read page walkthrough for real UI journeys and browser evidence.

    line 86
  9. Distinctive:

    Read test record for functional plans, defects, coverage and recovery.

    line 87
  10. Distinctive:

    Read roles and performance only for relevant multi-user, authorization or performance requirements.

    line 88
  11. Distinctive:

    Read result contract when consuming results or returning evidence.

    line 89
  12. Distinctive:

    Read change preservation when code deletion or instruction changes participate.

    line 90
  13. Distinctive:

    Read performance measurement for scoped comparative performance acceptance.

    line 91

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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.

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

Dev Tests

Entry Gate

Own a bounded test-coverage and evidence decision for a declared feature, change, or journey. For existing projects, default to deriving and running cases from current code, interfaces and available page observations; a complete PRD is not required. Record conflicts or unverifiable business meanings for later human confirmation, continue independent checks, and withhold only dependent business verdicts or unsafe operations. During product/design work, identify risks, testability gaps and provisional checks without waiting for implementation. Resolve applicable authorities, their acceptance status, target basis, environment and permission before execution; only settled business oracles support business-acceptance verdicts; baseline regression and provisional exploration retain their narrower meaning. Reuse the project test plan and its IDs. When asked to execute functional acceptance, run the reachable authorized cases; a checklist alone does not complete the run.

Route Map

Request conditionReadResult
AI-generated Playwright journeys or test healing applyAI E2E acceptanceFrozen cases, bounded authoring handoff and native-runner evidence
Design early checks or reconcile risk-based coverage against contractscoverageRequirement-to-layer matrix and qualified gaps
Run existing checks, rerun failures, or choose regression scopeexecutionReproducible scoped runs and evidence-bound outcomes
Execute a functional acceptance campaign or resume its case inventoryfunctional workflowReal journeys, correlated results, retest and cleanup
API or business-state behavior participatesAPI and logicWire, business status and independent persistence assertions
Pages, UI or clicks participatepage walkthroughReachability, real interaction and visible outcomes
Record functional cases, defects or recoverytest recordFinite coverage, case evidence and recovery queue
Deletion, moves/renames, identifiers, enforcement rules or instructions changechange preservationOriginal acceptance, positive controls, consumer regressions and qualified comparison
Scoped performance acceptance or before/after comparison participatesperformance measurementComparable metrics, protected journeys and qualified verdicts
Role isolation, concurrent users or load scenarios applyroles and performanceAuthorized scenarios, metric interpretation and comparable measurements
Compose a result for review or another ownerresult contractVersioned basis, coverage, evidence and handoff

Invariants

  • Derive user-facing copy and feedback type/timing expectations from the original request and accepted Product/UI decisions, not candidate output. Exercise the affected state; distinguish action dispatch from progress/completion. Candidate specs, tests and green counts cannot approve an unsupported presentation change.
  • Product behavior belongs to product-spec, visual behavior to ui-spec, and wire semantics to api-spec. Test ambiguity cannot silently decide those contracts.
  • Select and run existing project tests within scope. Missing tests, harness changes, fixtures requiring source edits, and fixes belong to the matching implementation owner. Browser page/DOM mechanics belong to ops-browser, desktop-client control to ops-client, and mobile device/app lifecycle or native UI to ops-device. These names are handoff targets, not automatic invocation or granted capabilities.
  • Read current project commands and their effects before execution. Use isolated test state and authorized resources; preserve safety stops, denied routes and scope limits. A failure does not authorize another identity, mechanism, or environment to bypass them.
  • Preserve passed, failed, blocked, not-run, and excluded separately. A mock, fixture, static check or lower layer proves only its observed boundary, never a real integration or complete E2E. Missing access is a blocker, not an exclusion.
  • Rebind evidence after requirement, UI, code, configuration or environment changes. Reuse unaffected evidence only with an explicit impact rationale; old green results cannot approve a new basis by default.
  • Select raw observations for the actual assertions and estimate output growth before scaling a capture matrix. Follow the test record for evidence sufficiency; storage retention and cleanup remain with the host owner.
  • Produce test evidence, not a competing task ledger or a release verdict. to-task owns durable work reconciliation, repo-review the fixed-basis review, and repo-delivery separately authorized Git actions. No deployment, installation, publication, production load, or implicit promotion follows from testing.
  • Consolidate delegated case results into the frozen inventory before selecting more work. An agent terminal state closes only its assigned execution, not another evidence layer or the run verdict.

Output Map

Return the result contract: frozen basis and authority versions, layer matrix, executed checks and evidence, failures and blockers, justified exclusions, remaining acceptance gaps and smallest owner-specific next action. State whether a first real E2E was observed and its exact boundary. Never flatten the matrix into “all tests passed” when required cells remain unresolved.

Reference Map

  • Read Playwright E2E only for coordinating generated browser-test acceptance.
  • Read coverage when deriving or reconciling the acceptance matrix.
  • Read execution before running checks or making regression claims.
  • Read functional workflow for functional campaigns and resumed case inventories.
  • Read API and logic for network, business status and persistence.
  • Read page walkthrough for real UI journeys and browser evidence.
  • Read test record for functional plans, defects, coverage and recovery.
  • Read roles and performance only for relevant multi-user, authorization or performance requirements.
  • Read result contract when consuming results or returning evidence.
  • Read change preservation when code deletion or instruction changes participate.
  • Read performance measurement for scoped comparative performance acceptance.
  • Maintainers only: read eval cases; do not load it during ordinary execution.

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.