agentleFS
bordenetbordenet/superpowers-plus/skills/engineering/requirements-validation/skill.md

requirements-validation

Skill8 starsChanged 6 months agoMITView on GitHub

Use when validating feature requirements before design or implementation. Tests each requirement for falsifiability, measurability, and independence. Detects contradictions and guides resolution without resolving silently.

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    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:

    Before design work — validate that requirements are testable and non-contradictory

    line 39
  2. Distinctive:

    When stakeholder requirements seem vague, compound, or conflicting

    line 40
  3. Distinctive:

    When acceptance criteria need to be derived from prose requirements

    line 41
  4. Distinctive:

    Before running the three tests, normalize requirements into a numbered list:

    line 45
  5. Distinctive:

    If the input is prose, extract discrete requirements and number them.

    line 48
  6. Distinctive:

    If the input is already numbered, preserve the numbering.

    line 49
  7. Distinctive:

    When two requirements conflict:

    line 88

≈ “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

Requirements Validation

Core principle: Every requirement must be testable. Contradictions must be surfaced, not silently resolved.

Wrong skill? Feature design → debate or brainstorming. Implementation planning → plan-and-execute. Validating code output → output-verification.

Announce at start: "I'm using the requirements-validation skill to validate these requirements."

Companion Skills

  • debate: Evaluating design options from requirements
  • brainstorming: Generating approaches that meet requirements
  • plan-and-execute: Implementing validated requirements

When to Use

  • Before design work — validate that requirements are testable and non-contradictory
  • When stakeholder requirements seem vague, compound, or conflicting
  • When acceptance criteria need to be derived from prose requirements

Input Contract

Before running the three tests, normalize requirements into a numbered list:

  • Format: R1: [requirement text], R2: [requirement text], etc.
  • If the input is prose, extract discrete requirements and number them.
  • If the input is already numbered, preserve the numbering.
  • Each R# must be a single, atomic requirement (split compound requirements).

The Three Tests

For EACH numbered requirement, apply all three:

1. Falsifiability Test

Question: Can you write a test that would FAIL if this requirement isn't met?

ResultAction
Yes — concrete test exists✅ Requirement passes
No — too vague to test❌ Rewrite to be specific. "Improve performance" → "Response time < 200ms at p95"
Partially — some aspects testable⚠️ Split into testable and non-testable parts

2. Measurability Test

Question: Is "done" a binary state (yes/no), not a gradient?

ResultAction
Binary — clear done/not-done✅ Passes
Gradient — "better", "improved", "enhanced"❌ Add a threshold. "Better error handling" → "All error paths return structured error with code, message, and recovery hint"

3. Independence Test

Question: Does this requirement conflict with any other requirement in the set?

ResultAction
Independent — no conflicts✅ Passes
Conflicts detected❌ Trigger contradiction resolution (below)

Contradiction Resolution

⛔ HARD GATE: Do NOT resolve contradictions silently. The stakeholder decides.

When two requirements conflict:

  1. State both requirements verbatim.
  2. State the contradiction explicitly: "R3 requires X, but R7 requires Y. These cannot both be true because [reason]."
  3. Propose resolution options:
    • Option A: Prioritize R3 (drop R7 or modify it)
    • Option B: Prioritize R7 (drop R3 or modify it)
    • Option C: Split into phases (R3 in v1, R7 in v2)
    • Option D: Merge into a new requirement that satisfies both constraints
  4. Record the decision:
    • Decision owner: [name or role — must be a real person who provided the decision. If unknown, write PENDING (ask user) and STOP.]
    • Chosen option: A / B / C / D
    • Rationale: [1 sentence]
  5. Do NOT proceed until recorded. Unresolved contradictions block Phase 2.

⛔ Never invent a decision owner. If no stakeholder has weighed in, the decision is PENDING. Do not fabricate a name or role to unblock yourself.

Output Format

## Requirements Validation Report

### Passed
- R1: [requirement] — Falsifiable ✅ Measurable ✅ Independent ✅
- R2: [requirement] — Falsifiable ✅ Measurable ✅ Independent ✅

### Failed (Needs Revision)
- R3: [requirement] — Measurability ❌ (gradient language: "improve")
  - Suggested revision: [specific, measurable version]

### Contradictions Found
- R4 vs R6: [description of conflict]
  - Resolution options: A / B / C / D
  - Decision owner: [name/role — or PENDING (ask user)]
  - Chosen option: [A/B/C/D — or PENDING]
  - Rationale: [1 sentence — or PENDING]

Common Failure Patterns

PatternExampleFix
Gradient language"Improve the UX"Add threshold: "Task completion time < 30s"
Compound requirements"Fast AND flexible AND secure"Split into 3 independent requirements
Implementation as requirement"Use Redis for caching"Restate as need: "Cache layer with <10ms reads"
Negative-only"Don't break existing behavior"State positive: "All existing tests pass"

Example

# Check that all requirements have acceptance criteria
grep -c "GIVEN.*WHEN.*THEN" requirements.md || echo "⚠️ Missing acceptance criteria"
# Verify testability: each requirement maps to at least one test
grep -rn "describe\|it(" test/ --include="*.ts" | wc -l

Failure Modes

FailureFix
Resolved contradictions silently without surfacing to userHard gate: ALL contradictions go to user with options, never resolve silently
Accepted vague requirements (fast, secure) as testableApply falsifiability test: if you cannot write a test that fails, it is not testable
Skipped independence test, creating hidden couplingCheck: can each requirement be implemented/verified without the others?
Validated requirements against assumptions instead of stated constraintsCross-reference ONLY what the user stated, not what you inferred

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.