The Agent Skills Directory

[SAFE]: No malicious instructions, prompt injections, or obfuscation techniques were identified across the skill files. The instructions are focused on professional project management and team leadership.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from external sources, including workspace files, team communication channels, and output from other agents. However, it incorporates strong defensive instructions:
Ingestion points: Data is read from the current workspace and inter-agent communication channels (SKILL.md, references/communication.md).
Boundary markers: The skill defines a structured "Assignment and return contract" in references/communication.md to delimit task instructions and results.
Capability inventory: The skill can read/write to the local .team/ directory and delegate tasks to subagents.
Sanitization: It explicitly requires agents to "Inspect findings," "Examine consequential evidence," and verify that artifacts actually show the claimed results before acceptance.
[DATA_EXPOSURE]: The skill includes explicit security warnings in references/team-memory.md to "Keep secrets and confidential business context out of reusable skill material" and only store necessary project information in the local folder.
[COMMAND_EXECUTION]: The skill utilizes standard GitHub CLI commands (gh api, gh pr diff) for coordinating parallel agents within a repository. These tools are used appropriately for managing development workflows and isolating agent worktrees.

team-lead