메뉴
Phase 6 of Eiffel Spec Kit. Generates adversarial tests, stress tests, and edge case tests. AI chain suggests attack vectors. Use with /eiffel.harden command.
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
직접 명령은 검토 Prompt를 거치지 않습니다. 실행하기 전에 소스를 확인하세요.
npx skills add https://github.com/simple-eiffel/eiffel-spec-kit --skill eiffel-harden명령은 한 줄로 유지됩니다. 복사하기 전에 가로로 스크롤해 전체 내용을 확인하세요.
로컬 사본을 원하시나요? SkillsMP에서 현재 제공할 수 있는 파일을 다운로드하세요.
SOC 직업 분류 기준
SKILL.md 표시 중
| name | eiffel-harden |
| description | Phase 6 of Eiffel Spec Kit. Generates adversarial tests, stress tests, and edge case tests. AI chain suggests attack vectors. Use with /eiffel.harden command. |
| allowed-tools | Read, Grep, Glob, Edit, Write, Bash, Task |
Purpose: Break the implementation with adversarial inputs, stress tests, and edge cases.
/eiffel.harden <project-path>
<project-path>/
├── .eiffel-workflow/
│ ├── prompts/
│ │ └── phase6-adversarial-review.md
│ └── evidence/
│ └── phase6-tests.txt
├── test/*.e (adversarial tests added here)
Before adversarial tests, strengthen contracts with Mathematical Model Library (MML).
Add simple_mml dependency to ECF:
<library name="simple_mml" location="$SIMPLE_EIFFEL\simple_mml\simple_mml.ecf"/>
Add MML model queries for collections:
feature -- Model Queries
items_model: MML_MAP [K, V]
-- Mathematical model of stored items.
do
create Result.make_empty
across internal_table as ic loop
Result := Result.extended (@ic.key, @ic.item)
end
end
Strengthen postconditions with frame conditions:
ensure
-- What changed
key_added: items_model.has (a_key)
value_set: items_model [a_key] = a_value
-- What did NOT change (frame condition)
others_unchanged: items_model.removed (a_key) |=| old items_model.removed (a_key)
MML Checklist:
|=| for equality comparisonsold items_modelCreate <project-path>/.eiffel-workflow/prompts/phase6-adversarial-review.md:
# Adversarial Test Suggestions
## Contracts
[PASTE src/*.e]
## Current Tests
[PASTE test/*.e]
## Suggest Tests For:
1. Boundary values (0, 1, MAX_INT, empty strings)
2. Capacity limits (full, zero capacity)
3. SCOOP concurrent access
4. State transitions in unexpected order
5. Resource exhaustion
User may submit prompt to external AI for adversarial suggestions.
Add tests like:
test_with_zero_capacitytest_same_key_repeatedlytest_stress_many_operationstest_concurrent_access_scoopCRITICAL: You MUST cd to the project directory before compiling. The EIFGENs folder is created in the current working directory, not where the ECF file is located. Compiling from the wrong directory pollutes other folders with build artifacts.
cd <project-path> && /d/prod/ec.sh -batch -config <project>.ecf -target <project>_tests -c_compile
./EIFGENs/<project>_tests/W_code/<project>.exe
ZERO WARNINGS POLICY: If the compiler reports obsolete call warnings or any other warnings, FIX THEM IMMEDIATELY. Do not note them and move on. Do not defer them. Fix them right now before proceeding.
ALL TESTS MUST PASS.
Why: This catches integration issues that don't appear in standalone compilation. A library with concurrency=scoop may compile fine but cause VUAR(2) errors in consumer libraries.
Create mock consumer project (temporary, can delete after test):
mkdir -p <project-path>/test_consumer
Create <project-path>/test_consumer/consumer_test.ecf:
<?xml version="1.0" encoding="ISO-8859-1"?>
<system name="consumer_test" uuid="[generate-new-uuid]">
<target name="consumer_test">
<root class="CONSUMER_TEST" feature="make"/>
<capability>
<concurrency support="scoop"/>
<void_safety support="all"/>
</capability>
<library name="base" location="$ISE_LIBRARY\library\base\base.ecf"/>
<library name="<project>" location="..\<project>.ecf"/>
<cluster name="src" location="."/>
</target>
</system>
Create <project-path>/test_consumer/consumer_test.e:
class CONSUMER_TEST
create make
feature
make
local
l_obj: <MAIN_LIBRARY_CLASS>
do
create l_obj.make
-- Use library types in SCOOP context
end
end
Compile consumer:
cd <project-path>/test_consumer && /d/prod/ec.sh -batch -config consumer_test.ecf -target consumer_test -c_compile
If VUAR(2) errors: Fix generic constraints in the library (add separate keyword).
Cleanup:
rm -rf <project-path>/test_consumer
Save to <project-path>/.eiffel-workflow/evidence/phase6-tests.txt:
# Phase 6 Hardening Evidence
# Project: <project-path>
# Date: [timestamp]
## MML Hardening
MML dependency added: [yes/no]
Model queries added: [count]
Frame conditions added: [count]
## SCOOP Consumer Integration
Consumer test created: [yes/no]
Consumer compilation: [PASS/FAIL]
VUAR(2) errors fixed: [count or N/A]
## Adversarial Testing
Adversarial tests added: [count]
Stress tests added: [count]
[Paste test output]
# Status: PASS / FAIL
This skill focuses ONLY on: <project-path>
DO NOT:
DO:
Example - Ecosystem Query:
Need ideas for stress testing patterns? Don't read all test files in ecosystem.
Instead: Task(subagent_type=Explore) → "What adversarial test patterns exist in simple_container tests?"
The sub-agent searches, summarizes, and returns ONLY what you need. Your context stays focused on this project.
Phase 6 COMPLETE: Hardening passed.
Project: <project-path>
All tests: PASS
Next: Run /eiffel.ship <project-path> for final checklist.