docs(feat-028): scrub market-positioning + competitor names from the spec

docs/features/README.md

@@ -43,7 +43,7 @@
 | **feat-009** | Observability — structured logging (JSON) + distributed tracing (OTel) + hook fan-out; vendor backends (Langfuse / Phoenix / Evidently / StatsD) slated for v0.2 | shipped (Python, OTel only) | 0.1 (framework + OTel — shipped), 0.2 (vendor backends) | both | `agentforge`, `agentforge-otel`, future `agentforge-langfuse`, `agentforge-phoenix`, `agentforge-evidently`, `agentforge-statsd` |
 | **feat-026** | Application config extension — reserved `app:` namespace + typed `app_as()` accessor (Phase 1), registered typed sections validated via the module-schema engine + entry points (Phase 2), pluggable config sources / separate files via `imports:` (Phase 3). Lets agents built on AgentForge reuse the config machinery (interpolation, layering, `--resolved`, uniform validation). Reported via #86 | shipped (all 3 phases → 0.3.0) | 0.3.0 | both | `agentforge-core`, `agentforge` |
 | **feat-027** | Embedded `GraphStore` — `KuzuGraphStore`, a zero-ops, file-backed, in-process graph driver (the graph analogue of the SQLite `MemoryStore`). Implements the locked `GraphStore` ABC and passes `run_graph_conformance`, so it is swap-compatible with Neo4j/SurrealDB; `path: .ckg` and the store exists — no server. Makes the whole graph + GraphRAG path testable offline. Drives the `agentforge-graph` code-graph dogfood | implemented (0.4) | 0.4 | python (TS deferred) | new `agentforge-memory-kuzu` |
-| **feat-028** | Durable execution + human-in-the-loop — checkpoint-and-resume so a run survives process death (resumes at the failed step, no double-spend / double-fire), plus approval gates that suspend before irreversible tool calls and resume on human approve/deny/edit. Promotes the feat-017 record/replay seam into a checkpoint seam (`__checkpoint` claims on a `MemoryStore`); budget snapshot + idempotency-seed restore for correctness; `execution:` / `human_in_the_loop:` config blocks; additive `Agent` kwargs. The control-plane capability clouds charge for, open + vendor-neutral | proposed | 0.5 | python (TS deferred) | `agentforge-core`, `agentforge`, checkpoint driver pkg(s) |
+| **feat-028** | Durable execution + human-in-the-loop — checkpoint-and-resume so a run survives process death (resumes at the failed step, no double-spend / double-fire), plus approval gates that suspend before irreversible tool calls and resume on human approve/deny/edit. Promotes the feat-017 record/replay seam into a checkpoint seam (`__checkpoint` claims on a `MemoryStore`); budget snapshot + idempotency-seed restore for correctness; `execution:` / `human_in_the_loop:` config blocks; additive `Agent` kwargs. Foundational execution infrastructure every production agent eventually needs | proposed | 0.5 | python (TS deferred) | `agentforge-core`, `agentforge`, checkpoint driver pkg(s) |
 
 ### Safety & security
 

docs/features/feat-028-durable-execution-and-hitl.md

@@ -38,9 +38,11 @@ durable state. Today `Agent.run` executes the entire reasoning loop inside a
 single `await self._strategy.run(state)` — if that coroutine dies, the run is
 gone, and there is no interception point to pause it for approval.
 
-This is precisely the control-plane capability the managed platforms charge
-for (Bedrock AgentCore, Temporal Cloud, Step Functions). Building it **open and
-vendor-neutral** is the strategic white space.
+It is foundational execution infrastructure: hard to implement correctly
+(checkpoint consistency, exactly-once side effects, budget-correct resume,
+suspend/resume control flow) and needed by essentially every production agent
+— which is exactly why it belongs in the framework rather than in each agent
+(see §2.5).
 
 ## 2. Why it must ship as framework