The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use the GitHub CLI (gh) to list pull requests and inspect file changes. These commands are used for legitimate project management tasks like identifying file conflicts and tracking progress.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from teammate agent responses and repository content, which constitutes an attack surface for indirect prompt injection.
Ingestion points: Responses from other agents via communication channels (terminal sessions, subagent tools, inter-agent inbox), project artifacts (code/documentation), and memory files within the .team/ directory.
Boundary markers: The skill incorporates several mitigation strategies, including instructions to distinguish between observed and claimed identities, verify evidence manually (inspecting the resulting artifact), and use a structured 'assignment and return contract' to define task scope.
Capability inventory: The agent has permissions to read/write files in the workspace (specifically the .team/ folder and project artifacts) and execute shell commands via project-relevant CLI tools like gh and herdr.
Sanitization: The instructions focus on high-level verification of results and evidence before accepting worker contributions, rather than technical sanitization of inputs.

team-lead