The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends using bunx ccrecall sync in references/task-management.md, which fetches the ccrecall package from the public npm registry at runtime without version verification.
[REMOTE_CODE_EXECUTION]: Execution of the ccrecall package via bunx constitutes dynamic loading and execution of remote code from a non-trusted source.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to synchronize analytics data, involving the bunx package runner.
[DATA_EXFILTRATION]: The ccrecall sync operation reads team configuration data from ~/.claude/teams/ and transmits it to an external service for analytics purposes.
[PROMPT_INJECTION]: The skill describes patterns where agents ingest untrusted codebase content to generate tasks, creating a surface for indirect prompt injection. Ingestion points: Agents are tasked with reading files such as src/auth/ or CI logs to gather context for further actions. Boundary markers: The skill recommends using a 'Worker preamble' (e.g., 'CONTEXT: You are a WORKER agent...') and file partitioning to restrict agent behavior. Capability inventory: Tools like TaskCreate and TeamCreate provide the ability to read/write files and execute delegated tasks. Sanitization: No technical sanitization or escaping of ingested file content is specified before it is interpolated into sub-agent prompts.

orchestration