agentleFS
System-Error-WorldwideSystem-Error-Worldwide/the-loop/.agents/skills/test/SKILL.md

test

Skill0 starsChanged 3 months agoMITView on GitHub

Independently attempt to falsify the done gate, persist reproducible evidence and open every defect that survives refutation. Use when an approved implementation or non-code asset must be verified independently before it can be called green.

179 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use agent-context-isolation, agentica-infrastructure and 10 more.

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.

The file

We split this skill into the rules it gives an agent

4 rules. 0 are common, 3 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 154 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use after Build, after Resolve returns an issue to independent verification, or whenever a done-gate claim needs falsification.

    line 19
  2. Distinctive:

    Read authority for checks and exact authority plus matching lease for evidence or issue writes.

    line 26
  3. Very specific:

    run ID, stage test, track and declared asset.

    line 46
  4. Distinctive:

    Every passed condition needs current persisted evidence with a non-empty procedure and result reference.

    line 55

“Common” is measured: kept by at least 6 in 10 of the 154 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

Purpose

Provide the complete bundled fallback for lifecycle.test. Judge the artifact against declared requirements and observed evidence, not the implementer's success claim.

Use when

Use after Build, after Resolve returns an issue to independent verification, or whenever a done-gate claim needs falsification.

Required inputs

  • Validated run ID, declared asset, track, current stage and exact done gate.
  • Requirement IDs, changed artifacts, environment and relevant prior evidence.
  • Current issue ledger, regression procedures and provider route records.
  • Read authority for checks and exact authority plus matching lease for evidence or issue writes.
  • Frozen budgets, stop controls and known unavailable capabilities.

Reject a test request whose acceptance conditions or observable environment cannot be identified.

Procedure

  1. Derive checks from the done gate, requirements, changed behavior, failure paths and track contract.
  2. Attempt to falsify every acceptance claim before accepting it.
  3. Exercise the normal path plus relevant boundaries, invalid inputs, permission denial, interruption, recovery and truthful unsupported behavior.
  4. Capture reproducible evidence with procedure, actor, time, environment, outcome and result artifact.
  5. Reproduce each candidate defect and try to refute it. Drop candidates that do not survive.
  6. For every surviving defect, open or update one issue with severity, blocking state, minimal reproduction, expected and observed behavior and regression procedure.
  7. Close an issue only when new passing regression evidence exists; preserve the original failure evidence.
  8. Report green only when all required evidence passes and no blocking issue remains open.

Output contract

Return:

  • run ID, stage test, track and declared asset;
  • checks attempted, coverage classification and exact environment;
  • reproducible evidence references for each requirement;
  • surviving defect and issue references;
  • unexecuted, denied, blocked and unverified checks;
  • gate state and next stage: resolve, close or a truthful halt.

Evidence gate

Every passed condition needs current persisted evidence with a non-empty procedure and result reference. Any missing, stale, failed, blocked, denied or unverified required check blocks green. A screenshot proves only its captured visual state.

Track requirements

For the code track, select justified build, static, unit, integration, security and behavior checks and preserve commands plus exit status. For the non-code track, test source quality, factuality, calculations, completeness, requested format, render output, audience fit and domain-specific review. Both code track and non-code track label bounded or sampled coverage explicitly.

Safety and authority

Evidence collection stays within current authority. A read-only test cannot mutate production or contact an external system. Issue and evidence writes use the runtime intent boundary. Provider failure is recorded before rerouting and unknown operations are never replayed.

Self-refutation

Attempt to reproduce each claimed defect and break each claimed pass. Check that a passing command actually exercises the requirement and that the observed artifact matches the inputs under review.

Halt conditions

Halt green promotion when a required check cannot run, evidence is stale or invalid, a blocking issue remains, authority is denied, the lease is lost, a stop is detected, the namespace changes, or the retry gate is exhausted. A reopened issue or three red Test/Resolve passes requires a human gate.

Core execution does not require network access or the source checkout. The links below are optional public documentation.

References

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.