The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the faf-cli tool globally via npm. This package is managed by a source not listed as a trusted vendor.
[REMOTE_CODE_EXECUTION]: The configuration examples utilize npx to download and execute the claude-faf-mcp package. This allows for the execution of remote code at runtime.
[COMMAND_EXECUTION]: Provides and uses various CLI tools (faf init, faf score, faf bi-sync) to perform file operations and system checks.
[PERSISTENCE]: The faf mcp install command modifies Model Context Protocol configuration files. This registers third-party tools as persistent agents that run in subsequent AI sessions.
[INDIRECT_PROMPT_INJECTION]: The skill processes external project data to create AI behavioral instructions, introducing an attack surface.
Ingestion points: Reads project identity, stack details, and architecture from .faf and other local files.
Boundary markers: There are no explicit instructions or delimiters to prevent the agent from being influenced by malicious text within the processed project files.
Capability inventory: The skill can execute shell commands and write to sensitive instruction files like .cursorrules and CLAUDE.md.
Sanitization: No sanitization logic is present to filter out potentially malicious prompt content from the ingested project DNA.

faf-expert