skills-quirkBLOCK

From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.

git clone https://github.com/quantumquirkxyz/skills-quirk.git
cd skills-quirk
# Node.js 18+ and Bash are required; CI uses Node.js 20.
node .agents/skills/platform/check-all.mjs

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: integration-playground
description: A playground environment for testing and developing quirk skills locally.
version: 1
capabilities:
  - create-skill
  - validate-skill
  - run-skill
  - export-skill
  - import-skill
inputs:
  - skill-template: Path to a skill template (optional)
  - test-case: Description of the test scenario
  - environment: Target environment (default: general)
outputs:
  - skill-result: The resulting skill object with validation evidence
  - test-report: Detailed test results
  - execution-log: Full execution log
stopCondition: The skill result is complete and validation passes
risk: medium
trustTier: 3
maxIterations: 10
---

# Integration Playground

## Contract
- Input: test case and environment specification
- Output: skill result with validation evidence
- Boundary: isolated from production code

## Process
1. Parse the test case.
2. Execute against the environment.
3. Collect output and validation evidence.
4. Report results.

## Guardrails
- Never write to the target repo outside the sandbox.
- Surface uncertainty and failures explicitly.

BLOCKgrade F · trust 40/100 Do not install this without reading the findings. The audit found something that could harm you or your machine.

What the source does

Filesystem

declared (5 observation(s))

Network

declared (4 observation(s))

Shell

declared (4 observation(s))

Dependencies

not all pinned

Secrets in source

none-found

Findings (25)

{ pattern: /eval\s*\(/, label: 'eval() usage' },

{ pattern: /eval\s*\(/, label: 'eval() usage' },

- Model access (API keys, rate limiting, retries, fallback models).

- **Access control:** IP allowlisting, password protection, preview deployment access.

- Safety and bias evaluation: toxicity, demographic parity, jailbreak resistance.

| prompt-injection | critical | ignore previous instructions, disregard above, role overrides (DAN, unrestricted), <<SYS>> / [INST] markers |

validate_chart.cpython-312.pyc

graphic-design-accessibility.zip

graphic-design-advertising.zip

graphic-design-data-viz.zip

graphic-design-editorial.zip

.agents/skills/implement

.agents/skills/implement-review-fixes

.agents/skills/platform/schemas/code-review-schema.json

.agents/skills/platform/schemas/review-pr-schema.json

.claude/skills/accessibility

category: 'network-exfil',

.generated-notes.md

.env.template

const helpers = new Function(

} from '../../skill-dev/skill-optimizer/scripts/proposal-policy.mjs';

} from '../../../../evals/run-behavioral.mjs';

import { validateFixture } from '../../../../evals/run-behavioral.mjs';

react, react-dom, @types/react, @types/react-dom, @vitejs/plugin-react, vite

- **Unauthorized access:** revoke credentials immediately, audit logs, notify security team.

Gates applied: instruction_override, no_behavioural_pass.