This directory is the single top-level home for official SkillRun example capsules.
These examples are self-contained and participate in Core validation or local release checks:
meeting_action_brief: product/adoption hero for meeting notes, action items, risks, follow-up copy, artifacts, and.skrpackaging.refund: contract/policy-boundary hero for approvals, structured errors, artifacts, run records, MCP exposure, and.skrpackaging.wecom_team_notice: local notification workflow with dry-run preview, approval boundaries, declared env, and markdown artifacts.commit_message_gate: Conventional Commits validation without auto-staging files.bounded_file_patcher: exact text replacement inside declared directories with patch artifacts.readonly_diagnostics_runner: named allowlist diagnostics without arbitrary shell strings.command_hello: Level 0 command adapter contract without a SkillRun SDK.
Generate the local Team Library hero catalog used by Desktop smoke and manual testing:
scripts/generate-desktop-hero-catalog.shThe generated catalog is written to target/desktop-hero-skr/catalog.json and
uses local file-source .skr packages with sha256 checksums. It includes
command_hello as the low-friction default smoke item and
meeting_action_brief as the product hero item when Python 3.13+ and pydantic
2.x are available for packaging.
The script copies tracked example files into target before running
skillrun manifest and skillrun pack, so generated manifests and packages do
not become source files. The catalog is for local inspection, install-plan, and
guarded apply testing only; it does not install dependencies, enable capsules,
mount clients, run actions, or mark packages trusted.
These examples also live under examples/ so the repository has one examples entry point. They may require external CLIs, credentials, or network access, and are not part of the mandatory Core CI path.
examples/lark_notice_sender: Feishu/Lark message sender throughlark-cli, constrained by chat_id allowlist, dry-run default, sensitive content checks, and audit artifacts.
Rules for optional external-tool examples:
- Each example must prove a single point: an external CLI or service can be constrained by a SkillRun contract.
- Preflight boundaries are mandatory: allowlists, dry-run defaults, or sensitive content checks.
- Audit artifacts are mandatory.
- The example must not be described as a general adapter, marketplace item, trusted package, or sandboxed execution.
- Optional external-tool examples should remain rare. If maintenance cost exceeds value, mark the example deprecated or remove it.