The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the prefactor CLI tool to provision agent credentials. The instructions specify running prefactor setup with the --json flag to retrieve environment configuration. This is standard behavior for a CLI-based bootstrap process.
[EXTERNAL_DOWNLOADS]: The skill recommends installing various NPM packages from the @prefactor scope (e.g., @prefactor/cli, @prefactor/langchain, @prefactor/ai). These are official resources provided by the vendor for SDK instrumentation.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to parse JSON output from a CLI command and map those values to environment variables. This creates a data ingestion surface where the agent processes external tool output.
Ingestion points: Output from prefactor setup --json (SKILL.md)
Boundary markers: The instructions specify mapping specific JSON fields (api_url, api_token, agent_id, agent_identifier) to predefined environment variable keys, which limits the scope of the data processing.
Capability inventory: The skill has the capability to write to local environment files and execute shell commands via the CLI.
Sanitization: The skill relies on the structured nature of the JSON output and specific field mapping to handle the data safely.
[SAFE_PRACTICE]: The instructions include specific security-conscious guidance, such as avoiding the manual creation of API tokens, warning against running prefactor login automatically, and explicitly instructing the agent not to commit prefactor.json to version control, as it may contain sensitive profile tokens.

bootstrap-existing-agent-with-prefactor-cli