The Agent Skills Directory

[Prompt Injection] (SAFE): No instructions designed to override safety filters or extract system prompts were detected. The skill uses instructional templates for standard project management tasks.
[Data Exposure & Exfiltration] (SAFE): No sensitive file access (e.g., SSH keys, cloud credentials) or network exfiltration patterns were identified. Network operations are limited to local Git commands.
[Unverifiable Dependencies & Remote Code Execution] (SAFE): The skill does not download or execute external scripts from the internet. All execution logic is contained within the provided local scripts.
[Persistence Mechanisms & Privilege Escalation] (SAFE): No attempts to gain administrative privileges or establish persistent access were found. All operations are limited to the project directory.
[Indirect Prompt Injection] (LOW): The skill ingests project specifications and tasks, which could theoretically contain malicious instructions. However, the risk is minimal given the manual review steps and local context.
Ingestion points: spec.md, plan.md, tasks.md
Boundary markers: Uses standard Markdown headers and designated template placeholders
Capability inventory: Local file system modification and Git branch management
Sanitization: Includes regex-based sanitization for branch names and task IDs

speckit-constitution-zh