Skip to main contentCrate shellfirm
Source - agent
- AI agent guardrails — non-interactive decision mode for agent-originated commands.
- audit
- Audit trail — optional local log of every intercepted command.
- blast_radius
- Blast radius — runtime impact computation for matched checks.
- checks
- Manage command checks
- context
- Runtime context detection.
- env
- Environment abstraction for testability.
- error
- Typed error types for shellfirm.
- llm
- LLM-powered semantic command analysis (feature-gated behind
llm). - mcp
- MCP (Model Context Protocol) server — exposes shellfirm as an MCP tool server.
- policy
- Project-level policies (
.shellfirm.yaml). - prompt
- Challenge prompts and the
Prompter trait for testability. - wrap
- PTY proxy for wrapping interactive programs (psql, mysql, redis-cli, etc.).
- AgentConfig
- Configuration for AI agent guardrails.
- CmdExit
- Config
- describe configuration folder
- LlmConfig
- Configuration for the optional LLM-powered command analysis.
- Settings
- Describe the configuration yaml
- SeverityEscalationConfig
- Configuration for severity-based challenge escalation.
- WrapperToolConfig
- Per-tool configuration for the
shellfirm wrap proxy. - WrappersConfig
- Configuration for the
shellfirm wrap PTY proxy.
- Challenge
- The user challenge when user need to confirm the command.
- DEFAULT_ENABLED_GROUPS
- value_set
- Set a value at a dot-notation path, creating intermediate mappings as
needed.