| mode | Execution mode:
- auto: Detect from event context (PR opened → Mode A, issue_comment → Mode B)
- governance: Mode A — policy checks on PR diffs
- run: Mode B — full plan → execute → persist
- validate: Mode B — run with self-correcting validation loop
- plan: Mode B — plan only, no execution
- status: Show current execution status
| no | auto |
| intent | Natural-language intent for Mode B (run/validate/plan). What should Rigorix do? | no | — |
| permission-mode | Permission mode for engine tool execution:
- read_only: Only read operations
- workspace_write: Read + write within workspace (CI default)
- dangerous_full_access: No restrictions
| no | workspace_write |
| policy-file | Path to policy.toml for Mode A governance. Loaded from BASE BRANCH to prevent tampering. | no | .rigorix/policy.toml |
| fail-on-violation | Mode A: Fail workflow on policy violations. Default false (warn-only for early adoption). | no | false |
| fail-on-action-error | Fail workflow if the action itself encounters an error. Default false (fail-open). | no | false |
| max-llm-calls | Maximum LLM API calls per execution | no | 50 |
| max-llm-tokens | Maximum LLM tokens per execution | no | 50000 |
| max-validation-iterations | Maximum validation loop iterations | no | 3 |
| max-retries | Maximum retries for transient API failures | no | 3 |
| retry-delay-ms | Base retry delay in milliseconds (exponential backoff with ±25% jitter) | no | 1000 |
| post-pr-comment | Post execution summary as a PR comment | no | true |
| profile | Configuration profile to use | no | — |
| provider | LLM provider (anthropic, openai, deepseek, custom) | no | anthropic |
| api-base-url | LLM API base URL (e.g. https://api.deepseek.com/v1 for DeepSeek, https://api.openai.com/v1 for OpenAI) | no | — |
| model | LLM model name (e.g. deepseek-chat, gpt-4o, claude-sonnet-4-20250514) | no | — |
| max-tokens | Maximum tokens per LLM call | no | 4096 |
| temperature | LLM temperature (0.0-1.0) | no | 0.3 |
| backend-api-key | API key for the audit/enforcement backend | no | — |
| backend-audit-url | URL for posting audit records (e.g. https://backend.example.com/audit) | no | — |