The Agent Skills Directory

[SAFE]: The skill implements a Goal Oriented Development (GOD) framework. It uses a series of sub-skills and local Python helper scripts to manage task states, plans, and implementations. No malicious behavior was detected.
[COMMAND_EXECUTION]: The skill utilizes shell commands to facilitate standard software development workflows. These operations are essential to the primary purpose of the skill and are performed using local environment tools.
Executes git for branch management, diffing, and staging changes (e.g., git checkout, git diff, git push found in sub-skills/implement/SKILL.md and sub-skills/_lib/parse_coverage.py).
Uses GitHub CLI (gh) for Pull Request creation and status verification (e.g., gh pr create in sub-skills/pack-up/SKILL.md).
Invokes local Python scripts within the sub-skills/_lib/ directory to perform deterministic parsing of markdown files and YAML frontmatter.
[SAFE]: The skill incorporates an 'Indirect Prompt Injection' surface analysis but follows best practices for mitigation.
Ingestion points: Fetches data from Jira issues and Figma designs using official MCP tools in sub-skills/spec/SKILL.md and sub-skills/code-like-me/SKILL.md.
Boundary markers: Employs a 'context blob' pattern (introduced in v10.4) that uses clear headers such as === status.md === to delineate pre-loaded files within the agent context.
Capability inventory: Has capabilities for file system modification, shell command execution via Bash, and interaction with GitHub through the CLI.
Sanitization: Uses dedicated Python scripts (e.g., parse_spec.py, parse_coverage.py) to extract requirements and acceptance criteria through regular expressions, minimizing the risk of untrusted data influencing agent logic directly.

god