Agent Core Extractor

Export the agent core from supported framework repositories into a small source-only zip for AI migration or cross-framework analysis. Use when the user want...

Security Scan

The skill's requests, files, and runtime instructions match its stated purpose (exporting agent-core files from local repos into a zip); it operates on local filesystem paths and does not request credentials or perform network exfiltration.

Purpose & Capability

The name/description, SKILL.md, README, and the included script all align: they detect repository framework signatures and copy a small set of source files into a staging directory and zip it. No unrelated credentials, binaries, or services are requested.

Instruction Scope

The runtime instructions are limited to running the bundled shell script which reads repositories under a configurable base directory (default ~/Documents/GitHub). This is in-scope for extracting agent-core files, but note the script will read arbitrary files within those repos (including any config files present) and will fail/exit if a specified repo is missing or unsupported.

Install Mechanism

No install spec is provided; this is an instruction-only skill with a bundled shell script. The script requires the zip binary (checked at runtime). No downloads or archive extraction from external URLs occur.

Credentials

The skill declares no required environment variables or credentials. The script uses HOME and PWD via shell expansion (normal) and accepts a --base-dir flag; there is no request for unrelated secrets or external tokens.

Persistence & Privilege

The skill is not always-enabled, does not modify other skills, and does not request elevated persistence. It performs filesystem reads and writes within the user-specified output directory only.

Assessment

This skill appears coherent and local-only, but be mindful before running it: point --base-dir at repositories you control, and run with a specific --repos list to avoid copying unexpected files. The exporter will copy config and prompt files from those repos — if those files contain secrets or provider keys, they will be included in the generated zip. Verify the zip contents before sharing, ensure the zip binary is installed, and run the script in a controlled environment (or on a copy of your repos) if you have sensitive data in your project trees.

AgentPearl Exporter

Use this skill to build a pure agent-core export from one or more supported repositories.

When To Use It

Use this skill when the user wants to:

• extract only the files that define an agent
• package agent-core files into a clean zip
• detect which supported framework a repo uses
• prepare source material for AI-driven migration into another agent framework
• study the identity, instruction, runtime, capability, multi-agent, or composition layers of an agent repo

Supported Frameworks

The bundled exporter detects these framework signatures:

• nanoclaw-ts-bootstrap
• nanobot-py-templates
• nullclaw-zig-bootstrap
• zeroclaw-rs-config-prompt
• openfang-rs-manifests
• codex-rs-builtins

Workflow

1. Confirm the target repositories exist under a common base directory.
2. Run scripts/export-agent-core-pack.sh.
3. Inspect the resulting zip, README.txt, and MANIFEST.txt.
4. If the user needs migration guidance, read references/AGENT_CORE.md to map the exported files into agent-core layers.

Commands

Export the default supported repositories under ~/Documents/GitHub:

./scripts/export-agent-core-pack.sh

Export a subset of repositories:

./scripts/export-agent-core-pack.sh --repos nanobot,openfang

Choose a base directory, output directory, and archive name:

./scripts/export-agent-core-pack.sh \
  --base-dir ~/Documents/GitHub \
  --output-dir ./out \
  --name agent-core-snapshot \
  --repos codex,nullclaw

Rules

• Keep the archive source-only.
• Include only files that directly define agent behavior.
• Exclude tests, build artifacts, dependency folders, unrelated app code, and target-framework metadata.
• If a repository does not match a supported signature, stop and report it instead of guessing.

References

• Read references/AGENT_CORE.md when you need the layer model for interpreting exported files.