Agents
Six lifecycle-specific agents with governed inputs, structured outputs, and mandatory human review for tapeout-critical actions.
Agent Network
Inputs, outputs, and lifecycle scope for each specialized agent.
Design acceleration and review intelligence
Inputs
4 sources
Outputs
4 types
Coverage closure and regression intelligence
Inputs
4 sources
Outputs
4 types
First-silicon debug orchestration
Inputs
4 sources
Outputs
4 types
Excursion detection and yield learning
Inputs
4 sources
Outputs
4 types
Field failure and RMA intelligence
Inputs
4 sources
Outputs
4 types
Institutional memory and engineering retrieval
Inputs
4 sources
Outputs
4 types
RTL Agent
Accelerates spec-to-RTL workflows, surfaces CDC/RDC risks, and provides continuous lint-aware review across large hierarchies.
Inputs
Outputs
Workflow
Lifecycle: Specification · RTL
RTL Agent
chipgpt / agents / rtl-agent
STATUS · Connected · awaiting engineering query
Awaiting engineering query...
→ Agent ready. All outputs require human approval.
Verification Agent
Bootstraps UVM environments, prioritizes coverage holes, and accelerates regression triage with historical closure patterns.
Inputs
Outputs
Workflow
Lifecycle: Verification · Validation
Verification Agent
Coverage Review — Interface Subsystem
REGRESSION RUN · build_8842 · subsystem review
Coverage gap identified in recovery-state bins
Pattern matched against prior tapeout closure log
→ Recommend directed tests aligned with historical closure approach
→ Traceability matrix references updated FMEDA artifacts
Sources: coverage.db, closure_log, verification_playbook
Bring-Up Agent
Structures lab failures, correlates register dumps, and proposes targeted debug experiments for bring-up teams.
Inputs
Outputs
Workflow
Lifecycle: Bring-Up · Validation
Bring-Up Agent
chipgpt / agents / bring-up-agent
STATUS · Connected · awaiting engineering query
Awaiting engineering query...
→ Agent ready. All outputs require human approval.
Yield Agent
Detects spatial defect patterns, correlates parametric drift to process tools, and accelerates yield RCA across lots.
Inputs
Outputs
Workflow
Lifecycle: Yield Learning · Production
Yield Learning Agent
Program Signal Review — Parametric Shift
YIELD LEARNING · parametric shift · program history review
Shift correlated with prior program learning record
Institutional context retrieved from earlier investigation
→ Hypothesis references validated cross-program pattern
→ Recommended review path preserves lifecycle continuity
Sources: yield_graph, program_history, knowledge_base
Failure Analysis Agent
Structures RMA narratives, correlates failure modes to errata, and accelerates FA workflows with institutional memory.
Inputs
Outputs
Workflow
Lifecycle: Production
Failure Analysis Agent
Field Escalation — Thermal Event Review
ESCALATION · thermal event · customer program review
Failure mode correlated with documented errata record
Historical engineering decision surfaced from prior FA cycle
→ Recommended FA path references validated internal playbook
→ Cross-program dependency flagged for related derivative
Sources: errata_db, fa_history, program_knowledge_graph
Knowledge Agent
Retrieves program history, surfaces prior engineering decisions, and maintains cross-lifecycle context for every specialized agent.
Inputs
Outputs
Workflow
Lifecycle: All lifecycle stages
Engineering Knowledge Retrieval
Query — CDC Requirements for Async FIFO Bridge
QUERY · CDC requirements for async FIFO bridge crossing
Retrieved relevant artifacts from program knowledge base
→ Design rule reference surfaced with source citation
→ Prior closure approach retrieved from institutional memory
Sources: design_rules, rtl_review_archive, cdc_closure_log