The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill's installation guide (references/sdd_install.md) contains a bash script designed for Linux environments that utilizes sudo dpkg to install the Git Credential Manager. Recommending commands that require administrative privileges increases the potential for privilege escalation attacks.
[EXTERNAL_DOWNLOADS]: The skill directs users and the agent to download the specify-cli tool and the Git Credential Manager from official repositories on GitHub. While the sources (e.g., github.com/github and github.com/git-ecosystem) are well-known and trusted, downloading and installing external software remains a critical security vector.
[COMMAND_EXECUTION]: The skill involves the execution of various local tools and scripts, including the specify command-line utility and several progress analysis scripts (e.g., scripts/phase_summary.sh, scripts/analyze-requirements.py). These scripts perform file system operations and text processing to support the development workflow.
[INDIRECT_PROMPT_INJECTION]: The skill functions by ingesting and analyzing user-generated content in files such as spec.md, tasks.md, and plan.md. These artifacts influence the agent's logic and the resulting implementation, providing a surface for indirect prompt injection if they contain malicious instructions. * Ingestion points: Specification files (spec.md), technical plans (plan.md), and task lists (tasks.md) located within the .speckit/features/ directory. * Boundary markers: None identified; the skill processes these documents as authoritative inputs. * Capability inventory: The agent executes shell commands, Bash scripts, and Python scripts. * Sanitization: No explicit content sanitization or safety filtering is performed.

sdd