Skip to content

dz import-ecc

dz import-ecc imports skill definitions from ECC (github.com/affaan-m/ECC) — a curated set of AI coding skills. DZ Harness Hub also ships an ECC-derived pack, canonicalized into the agentskills.io SKILL.md format. This command fetches SKILL.md files from the ECC repo (or a local clone), maps their frontmatter, deduplicates against your existing skills, and writes the result to an output directory.

Terminal window
# Import from the ECC GitHub repo (default)
dz import-ecc
# Import from a local ECC clone
dz import-ecc --local-path /path/to/ECC
# Import only the first 20 skills
dz import-ecc --limit 20
# Select specific skills by ID
dz import-ecc --select agentshield-scan,code-review,security-audit
# Write to a custom output directory
dz import-ecc --output ./my-ecc-skills
# Force overwrite skills that already exist
dz import-ecc --force

By default, imported skills are written to ./imported-ecc-skills/. Install them into a target with dz init --target claude-code --skills-dir ./imported-ecc-skills.

ECC is the GitHub repository affaan-m/ECC, a curated collection of AI agent skills. Each skill is a SKILL.md file with structured frontmatter. DZ Harness Hub re-ships selected skills as the skills-ecc pack.

The agentshield-scan skill (sourced from ECC) wraps npx ecc-agentshield scan, which runs 170 rules across 10 security categories.

ECC skills use their own frontmatter schema. dz import-ecc maps the fields it reads into agentskills.io format. The importer parses the name and description from the ECC frontmatter and synthesizes the rest:

Source agentskills.io Field Notes
name name Parsed from ECC frontmatter
description description Parsed from ECC frontmatter (multi-line collapsed)
(synthesized) trust_tier: 0 All ECC imports are hardcoded to tier 0
(synthesized) trust_tier_label "Community (imported from ECC)"
(synthesized) source https://github.com/affaan-m/ECC/tree/main/skills/<name>

Imported skills always receive trust_tier: 0 until you manually promote them. This signals that the skill has not been locally reviewed for your project’s security context.

Example: ECC Input → agentskills.io Output

Section titled “Example: ECC Input → agentskills.io Output”

ECC source (agentshield-scan/SKILL.md):

---
name: agentshield-scan
description: Run AgentShield security scanner (170 rules, 10 categories)
origin: ECC
tools: Read, Write, Bash
---

After dz import-ecc:

---
name: "agentshield-scan"
description: >
Run AgentShield security scanner (170 rules, 10 categories)
trust_tier: 0
trust_tier_label: "Community (imported from ECC)"
source: "https://github.com/affaan-m/ECC/tree/main/skills/agentshield-scan"
---

Before importing, dz import-ecc checks your existing .claude/skills/ directory. If a skill with the same ID already exists:

  • Without --force: the existing skill is skipped (counted as a duplicate)
  • With --force: the skill is re-imported and overwritten
Terminal window
# Import with overwrite of duplicates
dz import-ecc --force
Flag Description
--local-path <dir> Use a local ECC clone instead of fetching from GitHub
--select <ids> Comma-separated list of skill IDs to import
--limit <n> Maximum number of skills to import (default: all)
--output <dir> Output directory for imported skills (default: ./imported-ecc-skills)
--force Overwrite existing skills with the same ID
Terminal window
# Import all ECC skills into ./imported-ecc-skills/
dz import-ecc
# Import selected security skills from a local clone
dz import-ecc --local-path ~/ECC --select agentshield-scan,codeql-fix
# Limit to the first 50, write to a custom directory
dz import-ecc --limit 50 --output ./ecc-50
# Re-import everything (refresh)
dz import-ecc --force

Install the imported skills into a target, then validate them:

Terminal window
# Install the imported skills into Claude Code
dz init --target claude-code --skills-dir ./imported-ecc-skills
# List installed skills
dz list
# Benchmark an imported skill (19 L0 checks)
dz benchmark ./imported-ecc-skills/agentshield-scan
# After review, promote trust tier manually
# (edit the SKILL.md trust_tier field)