| Artifact | Structured document (PRD, RFC, ADR, etc.) stored in LanceDB |
| R_eff | Reliability score = min(evidence_scores). Weakest link, never average |
| Evidence | Test result, benchmark, or audit finding that supports/weakens a decision |
| Congruence Level (CL) | How close evidence context is to decision context. CL3=same, CL0=opposed |
| Depth | Rigor level: Tactical, Standard, Deep, Critical |
| Pipeline | Artifact creation sequence (e.g., PRD → RFC for Standard) |
| ADI | Abduction → Deduction → Induction reasoning cycle |
| Blind Spot | Active decision with zero linked evidence |
| Orphan | Artifact with no links to any other artifact |
| Stale | Artifact past valid_until date. Score drops to 0.1 |
| Superseded | Terminal state — replaced by newer artifact |
| Deprecated | Terminal state — no longer relevant |
| Validation Gate | Quality check before artifact activation (30+ rules) |
| Adversarial Review | Reviewer MUST find problems. 0 findings = re-review |
| PRD | Product Requirements Document — what & why |
| RFC | Request for Comments — how to build |
| ADR | Architecture Decision Record — why this approach |
| Epic | Groups PRDs, RFCs, ADRs into initiative |
| Spec | Specification — API contracts, data models |
| Problem | Signal with context — bug, risk, observation |
| Solution | 2-3 variant comparison with weakest-link scoring |
| Note | Micro-decision, auto-expires in 90 days |
| Refresh | Re-evaluation of stale artifact |
| Evidence Decay | TTL-based score degradation. Expired = 0.1 |
| F-G-R | Formality, Granularity, Reliability — quality dimensions |
| FPF | First Principles Framework — reasoning methodology |
| MCP | Model Context Protocol — AI agent integration |
| Route | Command that suggests depth + pipeline for a task |
| Projection | Markdown file generated from LanceDB record |
| Workspace | .forgeplan/ directory with config + storage |
| LanceDB | Embedded database for artifacts + vector embeddings |
