Agent-Skills.md

Agent Skills: Skill: cortex-toolkit

Unified toolkit for CORTEX operations including CAS integrity verification, summary generation, and LLM packer smoke testing.

UncategorizedID: Reneromero08/agent-governance-system/cortex-toolkit

Repository

Reneromero08/agent-governance-system
Reneromero08License: NOASSERTION

Install this agent skill to your local

pnpm dlx add-skill https://github.com/Reneromero08/agent-governance-system/tree/HEAD/CAPABILITY/SKILLS/cortex-toolkit

Skill Files

Browse the full folder contents for cortex-toolkit.

Download Skill

Loading file tree…

CAPABILITY/SKILLS/cortex-toolkit/SKILL.md

Skill Metadata

Name
cortex-toolkit
Description
"Unified toolkit for CORTEX operations including CAS integrity verification, summary generation, and LLM packer smoke testing."

required_canon_version: >=3.0.0

Skill: cortex-toolkit

Version: 2.0.0

Status: Active

Purpose

Unified toolkit for CORTEX operations including CAS integrity verification, summary generation, and LLM packer smoke testing.

Note: build and verify_system1 operations removed in v2.0 - semantic search is now handled by the cassette network (NAVIGATION/CORTEX/cassettes/).

Trigger

Use when performing any CORTEX-related operation:

  • Verifying CAS blob integrity
  • Generating section summaries
  • Running LLM packer smoke tests

Operations

| Operation | Description | |-----------|-------------| | verify_cas | Check CAS directory integrity (SHA-256 hash verification) | | summarize | Generate deterministic section summaries | | smoke_test | Run LLM Packer smoke tests |

Inputs

input.json fields:

Common:

  • operation (string, required): One of verify_cas, summarize, smoke_test

For verify_cas:

  • cas_root (string, required): Path to CAS root directory

For summarize:

  • record (object): SECTION_INDEX-style record with section_id, heading, etc.
  • slice_text (string): Exact section slice text

For smoke_test:

  • scope (string): ags or lab
  • out_dir (string): Output directory under MEMORY/LLM_PACKER/_packs/
  • mode (string): full or delta
  • profile (string): full or lite
  • combined (bool): Generate FULL/ outputs
  • stamp (string): Timestamp stamp for outputs
  • split_lite (bool): Generate LITE/ outputs
  • zip (bool): Generate zip archive
  • emit_pruned (bool): Generate PRUNED/ output

Outputs

output.json fields vary by operation:

For verify_cas:

  • status (string): success or failure
  • total_blobs (int): Number of blobs checked
  • corrupt_blobs (array): List of corrupt blob details
  • cas_root (string): Resolved CAS path

For summarize:

  • safe_filename (string): Safe filename for summary
  • summary_md (string): Generated summary markdown
  • summary_sha256 (string): Hash of summary

For smoke_test:

  • pack_dir (string): Output pack directory
  • stamp (string): Pack timestamp
  • verified (array): List of verified files
  • emit_pruned (bool): Whether pruned output was generated

Constraints

  • Writes only to allowed output roots via GuardedWriter
  • Deterministic: sets PYTHONHASHSEED, uses git timestamps
  • No network access
  • Read-only operations for verify_cas
  • No LLM calls for summarize operation

Fixtures

  • fixtures/basic/ - Basic operation tests for each operation type

required_canon_version: >=3.0.0