The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute various shell commands such as cat and grep to verify environment variables and plugin availability in ~/.claude/settings.json. It also provides a Python snippet to programmatically edit internal configuration files (~/.claude/teams/[팀이름]/config.json) if the automated team deletion process fails.
[PROMPT_INJECTION]: The skill exhibits vulnerability to indirect prompt injection due to its design of orchestrating data flow between subagents based on unvalidated external content.
Ingestion points: The skill accepts a user-provided topic ([주제]) and reads local project files to be processed by analyst and developer subagents as described in SKILL.md and references/team-patterns.md.
Boundary markers: The orchestration instructions do not implement explicit delimiters or boundary markers (such as XML tags or specific 'ignore' directives) to isolate ingested data from the agent's control instructions.
Capability inventory: The skill has the capability to spawn multiple subagents (analysts, critics, and Codex), read/write files, and execute arbitrary shell commands through the 'deployer' role template.
Sanitization: No input sanitization or escaping mechanisms are present to prevent malicious instructions within the analyzed data from influencing the orchestrator or its subagents.

peach-team-analyze