Fit and Limitations
Sigil is built for teams letting coding agents open or revise production pull requests. It is strongest when the risky part of the workflow is not code generation itself, but deciding whether an agent-authored change deserves to reach main.
Good fit
Section titled “Good fit”- You use Claude Code, Cursor, Devin, or a custom coding agent on real PRs.
- Your current alternative is visible tests, normal CI, and human-only PR review.
- You can express important behavior as markdown specs and executable scenarios.
- Your service can run in Docker or a reproducible CI environment.
- You want signed ALLOW, REVIEW, or BLOCK decisions with an audit trail.
Poor fit
Section titled “Poor fit”- You cannot deploy the service or a baseline in CI.
- You need broad static code review rather than behavior evaluation.
- You cannot safely provide test data, service secrets, or judge-provider configuration.
- You need a Windows-native install path today.
- You want to auto-merge before a service has accumulated clean evaluation history.
Operational assumptions
Section titled “Operational assumptions”Sigil assumes each protected service has explicit scenario ownership, a configured trust policy, and a ledger that can sync before an ALLOW decision. Any failure mode should downgrade to REVIEW or BLOCK.
Support
Section titled “Support”For fit questions or platform support, email info@runsigil.com. Include your CI provider, service runtime, agent workflow, and the command you want Sigil to gate.