The Agent Skills Directory

[SAFE]: The skill defines a structured development lifecycle. While it has extensive capabilities (interacting with GitLab, ArgoCD, and local files), these are aligned with its stated purpose as a development workflow orchestrator.
[COMMAND_EXECUTION]: The skill invokes various specialized tools (e.g., gitlab-mr, argocd, prometheus) to manage the R&D process. These actions are triggered based on the workflow state and require explicit human confirmation at each step, mitigating the risk of unauthorized execution.
[DATA_EXFILTRATION]: The skill reads project-level metadata (architecture docs, Makefiles, package configs) to generate documentation and a workspace summary. This data processing is local to the project environment.
[PROMPT_INJECTION]: The skill ingests external content via the lark-mcp tool and local project documentation.
Ingestion points: lark-mcp (external Feishu/Lark document links) and project-level files like docs/architecture/overview.md or package.json.
Boundary markers: No explicit delimiters or instructions are used to separate external content from internal prompts.
Capability inventory: The skill has broad capabilities including file writing (project initialization), merge request management (gitlab-mr), and deployment orchestration (argocd).
Sanitization: No explicit sanitization of ingested content is mentioned.
Mitigation: The workflow's design—requiring manual confirmation ([y/N]) before any step is finalized or progressed—provides a robust human-in-the-loop defense against potential injection attacks.

dev-workflow