AgentSkillsCN

cognitive-order-validator

通过结构性技能，强化认知层级关系——从“架构”到“逻辑”，再到“执行”。

SKILL.md
--- frontmatter
name: cognitive-order-validator
description: Structural skill to enforce the cognitive hierarchy (Architecture > Logic > Execution).

Cognitive Order Validator Skill

Status: Operational
Skill Category: Validation (Structural)

1. Skill Purpose

The cognitive-order-validator enforces the strict separation of concerns defined in the architectural_invariants.md. It ensures that "Thinking" layers (Signals, Narratives) do not perform "Acting" (Execution, Adapters) or import their dependencies.

2. Invocation Contract

Standard Grammar

code
Invoke cognitive-order-validator
Mode: <VERIFY | DRY_RUN>
Target: <path/to/src_dir>
ExecutionScope:
  mode: all
Options:
  enforce: <warnings | errors>

3. Supported Modes & Selectors

  • •VERIFY: Scan target files and return PASS/FAIL based on layer interaction rules.
  • •DRY_RUN: Perform the scan but return success even if violations are found (used for auditing legacy code).

4. Hook & Skill Chaining

  • •Chained From: Invoked as a Post-Execution Hook for implementation tasks.
  • •Chained To: Reports are consumed by the drift-detector.

5. Metadata & State

  • •Inputs: Source code, architectural_invariants.md.
  • •Outputs: Error report detailing illegal imports or keyword leaks.

6. Invariants & Prohibitions

  1. •Thinking != Acting: Pure logic files must not contain execution side-effects.
  2. •Unidirectional Flow: Upper layers cannot be imported by lower layers.
  3. •Market Logic Isolation: Trading primitives are forbidden in the Regime/Narrative layers.

7. Example Invocation

code
Invoke cognitive-order-validator
Mode: VERIFY
Target: src/narratives/
ExecutionScope:
  mode: all
Options:
  enforce: errors