TLM Forge vs GitHub spec-kit
GitHub spec-kit brings spec-driven development to your agent. TLM Forge shares the spec-first idea and adds what spec-kit leaves to you: adversarial review and a gate that blocks shipping until critical issues hit zero.
| Capability | TLM Forge | GitHub spec-kit |
|---|---|---|
| Where TLM Forge is ahead | ||
| Blocks the merge until every critical issue is resolved | ✓ | — |
| Adversarial reviewers — a threat-modeler at design, a red-team on the diff | ✓ | — |
| Ship or no-ship is a scored gate, not an opinion | ✓ | — |
| Enforced test-driven development with full-suite regression evidence | ✓ | — |
| Diagram and goal-contract sign-off before any code is written | ✓ | — |
| The basics — both cover these | ||
| A spec and a plan written before the code | ✓ | ✓ |
| A structured, staged workflow | ✓ | ✓ |
| Durable project context carried forward | ✓ | ✓ |
| Only in TLM Forge | ||
| Every behavior claim must cite file and line (code-grounding) | ✓ | — |
| Cross-project, team-wide memory | soon | — |
| A reproducible audit trail from spec to evidence | ✓ | — |
Why it works this way
Structure vs enforcement
A process you can skip is not a guarantee. Other tools hand the agent steps to follow; if it skips one, nothing stops it. TLM Forge makes the process mechanical — a commit-gate hook refuses the merge until every unresolved critical is gone.
Adversarial, not cooperative, review
Cooperative review assumes the work is basically right. TLM Forge launches reviewers in fresh context whose job is to attack it: a threat-modeler at design time, a red-team on the diff. A different mandate, and different blind spots than the agent that wrote the code.
A scored gate, not a vibe
Every finding is structured and scored. "Looks good" carries no weight — the gate is a number, and it has to read zero unresolved criticals before the work can ship.
Tests that are receipts
TLM Forge enforces test-driven development: a failing test before the code, then the full suite green with the run captured as evidence. "Tests pass" becomes a receipt you can reproduce, not a sentence you have to trust.
Built for Claude Code, on purpose
TLM Forge runs natively on Claude Code’s hooks, sub-agents, and MCP — the exact machinery that lets it block a merge instead of merely advising one. The depth is what makes real enforcement possible.