You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Parent epic: #157. Design anchors: DD-14 and DD-17.
Objective
Expose the corpus through stable human and agent interfaces without granting agents approval, source-snapshot mutation, lock replacement, or baseline-overwrite authority.
Human CLI scope
corpus source add|inspect backed by the M3 source store and transition validator.
Human-authorized source capture and canonical sources.lock writing without silent overwrite.
corpus ingest backed by M4 ingesters and drift contracts.
corpus validate, queue, context, review, trace, and render workflows as their backing contracts become available.
Complete PICS form modeling and rendering with requirement/profile bindings.
MCP and agent scope
Read-only corpus validation, source-context, extraction-queue, requirement-context, and trace tools.
Restricted append-only candidate-proposal API backed by M2.
No MCP operation for approval, review mutation, source acquisition/mutation, lock replacement, or baseline overwrite.
Surface contract
Register every new command/tool in KNOWN_SURFACES with matching HLR coverage.
Preserve JSONL single-terminal and stdout-strict invariants.
Preserve crate-root API discipline.
Add parity tests across CLI, MCP, doctor, and underlying core semantics.
Make human-vs-agent authority visible through available typed capabilities, not a caller-supplied actor string.
Dependencies
M2 lifecycle, review, and proposal authority: complete.
This is a milestone tracker. Source commands and lock writing, ingestion commands, validation/context/queue/review families, MCP read tools, proposal exposure, and PICS rendering each require dedicated child issues and independently revertible PRs.
Done when
Humans can manage and review the corpus through explicit authorized workflows; agents can inspect and append candidates only; PICS output is complete; every surface is trace-covered; and approval/source/baseline authority remains unavailable to agents by construction.
Parent epic: #157. Design anchors: DD-14 and DD-17.
Objective
Expose the corpus through stable human and agent interfaces without granting agents approval, source-snapshot mutation, lock replacement, or baseline-overwrite authority.
Human CLI scope
corpus source add|inspectbacked by the M3 source store and transition validator.sources.lockwriting without silent overwrite.corpus ingestbacked by M4 ingesters and drift contracts.corpus validate,queue,context,review,trace, andrenderworkflows as their backing contracts become available.MCP and agent scope
Surface contract
KNOWN_SURFACESwith matching HLR coverage.Dependencies
Delivery discipline
This is a milestone tracker. Source commands and lock writing, ingestion commands, validation/context/queue/review families, MCP read tools, proposal exposure, and PICS rendering each require dedicated child issues and independently revertible PRs.
Done when
Humans can manage and review the corpus through explicit authorized workflows; agents can inspect and append candidates only; PICS output is complete; every surface is trace-covered; and approval/source/baseline authority remains unavailable to agents by construction.