agentleFS
mageyukimageyuki/skills/design-pressure-test/SKILL.md

design-pressure-test

Skill1 starsChanged 2 months agoMITView on GitHub

Use when an approved design is explicitly requested for pressure testing, or is architectural, high-risk, or uncertain with material untested assumptions.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 5 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If either required input is absent, name the missing input and stop rather than inventing approval, risk, or assumptions.

    line 23
  2. Distinctive:

    Test each material assumption.

    line 31
  3. Very specific:

    Treat supplied passing checks as bounded evidence rather than demanding invented checks.

    line 38
  4. Very specific:

    Every material assumption has adequate bounded support: CLEAR.

    line 42
  5. Distinctive:

    No material blocker remains within the assumptions, scope, and evidence tested.

    line 61
  6. Distinctive:

    Keep challenges focused. Do not accumulate speculative blockers or reward criticism for being broader or harsher.

    line 75
  7. Distinctive:

    Do not reopen early intent discovery, edit the specification, redesign it, approve revisions, self-approve a replacement, prescribe …

    line 77

≈ “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.

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

Design Pressure Test

Purpose

Pressure-test the material assumptions in an approved design, report one bounded outcome, and stop.

Core principle: Challenge assumptions, not people. Test only what could materially change the design decision.

Entry contract

Begin only when both are supplied:

  1. An approved design or specification.
  2. Either a direct request from the current user, or an architectural, high-risk, or uncertain design with a material untested assumption.

An approved design alone is not a trigger. Authority or deadline pressure from someone else is not a direct user request.

If either required input is absent, name the missing input and stop rather than inventing approval, risk, or assumptions.

Recipe

  1. Identify only material assumptions.
    • Include assumptions about value, feasibility, safety, integrity, dependencies, migration, rollback, or operations only when failure could require a design change or prevent responsible assessment.
    • Ignore immaterial details and already-supported claims except to record why their bounded evidence is adequate.
    • Routine correctness and implementation details belong to implementation or code-review testing unless the supplied record establishes a specific design-changing risk; missing low-level test artifacts alone do not establish that risk.
  2. Test each material assumption.
    • State one strongest focused challenge.
    • State the concrete observation that would falsify the assumption.
    • State the consequence if it fails.
  3. Assess the supplied record.
    • Separate supplied facts and observations from inference and missing evidence.
    • Compare the actual evidence with the falsifying observation. Never invent tests, results, guarantees, or consent.
    • Treat supplied passing checks as bounded evidence rather than demanding invented checks. If no material unsupported assumption remains after the filter, select CLEAR within the stated scope.
  4. Select exactly one outcome.
    • Contradicted material assumption or necessary design change: REVISION_REQUIRED.
    • Necessary evidence is unavailable, so responsible assessment cannot finish: BLOCKED.
    • Every material assumption has adequate bounded support: CLEAR.
  5. Return the five-field report below and stop.

Report

  • Outcome: CLEAR, REVISION_REQUIRED, or BLOCKED.
  • Assumptions tested: For each material assumption, include the focused challenge and falsifying observation.
  • Evidence and observations: Distinguish supplied facts, observations, inference, and gaps.
  • Consequence: State what follows if the challenged assumption fails.
  • Required revision or blocker: Name the defective assumption or exact necessary evidence; use "None within the stated scope" for a clear bounded result.

For REVISION_REQUIRED, name what assumption or design claim must change. Do not supply a replacement design, code, API choice, or approval procedure.

For BLOCKED, name the unavailable evidence and why it is necessary. A justified block is a responsible result, not a failure to perform the test.

Outcomes

CLEAR

No material blocker remains within the assumptions, scope, and evidence tested. This is not blanket production approval.

REVISION_REQUIRED

Supplied evidence contradicts a material assumption, or the consequence requires the approved design to change.

BLOCKED

A material assumption cannot be responsibly assessed because necessary evidence is unavailable.

Boundaries

Evaluate the supplied approved design now; do not substitute a future test plan, new script, or new schema.

Keep challenges focused. Do not accumulate speculative blockers or reward criticism for being broader or harsher.

Do not reopen early intent discovery, edit the specification, redesign it, approve revisions, self-approve a replacement, prescribe implementation choices, or choose or invoke the next process.

Return the report to the design owner and stop.

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.