The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download and run the 'claude-flow' package from the NPM registry. This package is maintained by an external repository ('ruvnet/claude-flow') that is not identified as a trusted source for this environment.
[REMOTE_CODE_EXECUTION]: The use of 'npx' commands (e.g., 'npx claude-flow sparc run') results in the execution of remote scripts within the local execution environment.
[COMMAND_EXECUTION]: The skill describes workflows where user-supplied 'task descriptions' are passed as arguments to shell commands. This pattern creates a potential vector for command injection if the task strings contain malicious shell metacharacters.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface due to its data ingestion capabilities.
Ingestion points: The 'researcher' mode fetches content from external websites via web search and fetch tools, while the 'reviewer' mode processes code and metadata from GitHub pull requests.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions to isolate untrusted external content from the agent's core logic.
Capability inventory: The skill has powerful capabilities including autonomous code generation ('coder' mode), multi-agent task orchestration ('orchestrator' mode), and persistent memory storage ('memory-manager').
Sanitization: No sanitization, filtering, or validation steps are described for data retrieved from external URLs or code reviews before it is integrated into the agent's context.

sparc-methodology