Skill 编排核心

Skill 编排核心 - 上下文管理、流程编排、质量保证

What to check before installing or running this skill: - Review the full implementation of WorkflowOrchestrator._executeStage (file was truncated in the review). Confirm it does not invoke external network endpoints, execute arbitrary shell commands, or call other skill binaries unexpectedly. - Ensure required Node dependencies (e.g., js-yaml) are installed in a controlled way. The package provides no install spec; avoid running ad-hoc npm installs from untrusted repos. - The scripts read/write files under the project directory and create ~/.hermes/... paths in examples. If you plan to run it, inspect paths referenced in templates and the SKILL.md to avoid accidental reads of sensitive files in your home directory. - The SKILL.md uses a 'hermes-orchestrate' CLI name not present in the bundle; templates show running the scripts via node using an expected install location. Confirm how the skill will be invoked in your environment (wrapper CLI vs direct node execution). - Run the skill in an isolated/test project (not on sensitive repositories) first to verify behavior and to confirm there are no hidden network calls or side effects. Why this is 'suspicious' (not 'malicious'): - The code you can see is coherent with the stated purpose and contains no obvious exfiltration or credential access, but there are operational inconsistencies (missing install/dependency info and a truncated execution path) that create uncertainty about runtime behavior. If the truncated _executeStage is benign, this skill is likely fine; if it invokes shell/network operations, that would raise higher concern.

Type: OpenClaw Skill Name: skill-orchestration-core Version: 1.0.0 The skill bundle provides a legitimate framework for managing AI agent workflows, context, and output quality. The included Node.js scripts (scripts/context-manager.js, scripts/workflow-orchestrator.js, and scripts/output-validator.js) perform standard file system operations to maintain state and configuration within a local .orchestration directory. The logic is transparent, well-documented, and lacks any indicators of data exfiltration, malicious execution, or prompt injection.