Add explicit cross-agent session resume

.claude-plugin/skill-assets.sha256

@@ -3,4 +3,4 @@ c5d0c26f28c9ee14092f9deaf24c98dd8bef49d971fef2b7a537ffb1ab9f2887  .claude-plugin
 aeee7a94671ceb306fe2d24c5acc9f2d96ad8a8e7410536566799eea6265f080  skills/engraphis-memory/SKILL.md
 055655db84af07561d002f0c69744313d8413c39f3e873f941f0fa0b1e76dc66  skills/engraphis-memory/references/CONVENTIONS.md
 9d090a03f5b3f36a34d91f66b72c3844591f6915755ac3a6c6ba5f1b16977de5  skills/engraphis-memory/references/SCOPING.md
-07de31349fc135895377cfcf852c490f732c19e83a627c311d724c5d2146dbb5  skills/engraphis-memory/references/TOOLS.md
+68cd0add0d1025d495da2163e7c462b76b5fd1a4ff86b527fee49a4d8d6ab29b  skills/engraphis-memory/references/TOOLS.md

BENCHMARKS.md

@@ -94,14 +94,14 @@ interpretation and do not count as additional benchmark-quality gains.
 ### Public numeric evidence registry
 
 Every exact public aggregate retained below comes from the checked-in, public-safe
-[`offline-fixtures-v128.json`](docs/benchmark-evidence/offline-fixtures-v128.json) artifact. Its
+[`offline-fixtures-v129.json`](docs/benchmark-evidence/offline-fixtures-v129.json) artifact. Its
 SHA-256 is
-`73f2d1a8cd6e2db070577582a2266f6605efc052800da17db9938f7a274bf755`, also recorded in the
+`686394aa5d69cd296360defc5170d956c85565a5282c7766dc6466ba4c6aa6a5`, also recorded in the
 adjacent `.sha256` file. The artifact contains no raw questions, answers, prompts, customer data,
 or per-record content fingerprints.
 
 The fixture-suite digest is
-`6e1be135db67964cfb48a2125ac846a9a420662b69ed4d7f61c21d1c87d971b2`. The artifact defines
+`8e2a562ba5d4c785c1e88481f9019446b43608109147cd7cd82597493004af35`. The artifact defines
 the digest algorithm and records the SHA-256 of every suite and dataset file. Each evidence ID
 also binds its exact command through `sha256(UTF-8 exact command)`:
 
@@ -123,10 +123,10 @@ Historical LoCoMo, graph, handoff, consolidation, and security figures remain pr
 source artifacts but are omitted from the current chart until each has a matching immutable,
 public-safe artifact. The chart labels coding outcomes, external datasets, and operational
 capacity as pending evaluation tracks rather than implying scores. Regenerate it with
-`python scripts/render_benchmark_report.py --report docs/benchmark-evidence/offline-fixtures-v128.json --output docs/images/context-efficiency.svg` after selecting the report to publish.
+`python scripts/render_benchmark_report.py --report docs/benchmark-evidence/offline-fixtures-v129.json --output docs/images/context-efficiency.svg` after selecting the report to publish.
 
 The companion examples are also generated from that artifact with
-`python -m scripts.render_benchmark_examples --report docs/benchmark-evidence/offline-fixtures-v128.json --output docs/images/evidence-backed-agent-examples.svg`.
+`python -m scripts.render_benchmark_examples --report docs/benchmark-evidence/offline-fixtures-v129.json --output docs/images/evidence-backed-agent-examples.svg`.
 The historical-to-executable mapping is in
 [`docs/BENCHMARK_CHANGE_COVERAGE.md`](docs/BENCHMARK_CHANGE_COVERAGE.md).
 

README.md

@@ -84,9 +84,9 @@ neither is an end-to-end question-answer score. Coding outcomes, external datase
 operational capacity remain separate pending evaluation tracks until their artifacts are selected.
 
 These values are evidence IDs `offline-chunking` and `offline-performance` in
-[`offline-fixtures-v128.json`](https://github.com/Coding-Dev-Tools/engraphis/blob/main/docs/benchmark-evidence/offline-fixtures-v128.json),
+[`offline-fixtures-v129.json`](https://github.com/Coding-Dev-Tools/engraphis/blob/main/docs/benchmark-evidence/offline-fixtures-v129.json),
 SHA-256
-`73f2d1a8cd6e2db070577582a2266f6605efc052800da17db9938f7a274bf755`.
+`686394aa5d69cd296360defc5170d956c85565a5282c7766dc6466ba4c6aa6a5`.
 [`BENCHMARKS.md`](https://github.com/Coding-Dev-Tools/engraphis/blob/main/BENCHMARKS.md#public-numeric-evidence-registry)
 records the matching suite digest, exact commands, and per-command config digests. The offline
 fixture registry intentionally excludes external, model-dependent, consolidation, productivity,

docs/MCP_CONTRACT.json

@@ -1,6 +1,6 @@
 {
   "schema": "engraphis-mcp-contract/v1",
-  "sha256": "7e36e055ca003f17fa67194e1c22cc0af0e3c98172228b60989f1206dd1a28d2",
+  "sha256": "09291ce708f7ea4123252cfb217fef8e41d7043093f68b777f86f4204ab3661a",
   "surfaces": {
     "classic": [
       {
@@ -3027,7 +3027,7 @@
           "readOnlyHint": false,
           "title": "Start a memory session"
         },
-        "description": "Open a session to group this work's memories and enable cross-session resume.\n\nCall this at the start of a task in a repo you've worked in before \u2014 if a previous\nsession for the same authenticated user and agent was ended with a summary or open\nthreads, they come back in ``bootstrap`` so you can resume without crossing another\nuser or agent's handoff boundary.\n\nExact retries are reused by default for the same ``(workspace, repo, authenticated\nuser, agent, goal)`` identity. Different users, agents, or goals start distinct\nsessions automatically, and ``force_new=true`` always branches another session.\nBecause that valid option creates a new row on every call, the tool as a whole is\nconservatively annotated as non-idempotent.\n\nReturns:\n    str: JSON ``{\"session_id\",\"workspace\",\"repo\",\"goal\",\"status\":\"active\",\"reused\",\n    \"bootstrap\":{\"summary\",\"open_threads\",\"outcome\"} or {} if there is no prior\n    session}``. Pass ``session_id`` to engraphis_remember and engraphis_end_session.",
+        "description": "Open a session to group this work's memories and enable cross-session resume.\n\nCall this at the start of a task in a repo you've worked in before \u2014 if a previous\nsession for the same authenticated user and agent was ended with a summary or open\nthreads, they come back in ``bootstrap`` so you can resume without crossing another\nuser or agent's handoff boundary. ``resume_from_session_id`` explicitly selects an\nended handoff from another agent; it requires the same authenticated user and exact\nworkspace/repo, and it fails closed instead of choosing a different session.\n\nExact retries are reused by default for the same ``(workspace, repo, authenticated\nuser, agent, goal)`` identity. Different users, agents, or goals start distinct\nsessions automatically, and ``force_new=true`` always branches another session.\nBecause that valid option creates a new row on every call, the tool as a whole is\nconservatively annotated as non-idempotent.\n\nReturns:\n    str: JSON ``{\"session_id\",\"workspace\",\"repo\",\"goal\",\"status\":\"active\",\"reused\",\n    \"bootstrap\":{\"summary\",\"open_threads\",\"outcome\"} or {} if there is no prior\n    session}``. Explicit resumes also include source timestamps and bounded usage\n    metadata. Pass ``session_id`` to engraphis_remember and engraphis_end_session.",
         "inputSchema": {
           "properties": {
             "agent": {
@@ -3064,6 +3064,20 @@
               "description": "Repo scope, if any.",
               "title": "Repo"
             },
+            "resume_from_session_id": {
+              "anyOf": [
+                {
+                  "maxLength": 200,
+                  "type": "string"
+                },
+                {
+                  "type": "null"
+                }
+              ],
+              "default": null,
+              "description": "Optional exact ended session to hand off from another agent. It must belong to this authenticated user and exact workspace/repo; it never falls back to a recent session.",
+              "title": "Resume From Session Id"
+            },
             "workspace": {
               "anyOf": [
                 {
@@ -3754,7 +3768,7 @@
           "readOnlyHint": false,
           "title": "Start or end a memory session"
         },
-        "description": "Start/resume a session or end it with its next-session handoff.",
+        "description": "Manage sessions and explicit handoffs.",
         "inputSchema": {
           "properties": {
             "action": {
@@ -3778,7 +3792,7 @@
             },
             "goal": {
               "default": "",
-              "description": "Goal; start returns bounded context.",
+              "description": "Optional goal.",
               "maxLength": 1000,
               "title": "Goal",
               "type": "string"
@@ -3822,7 +3836,7 @@
             },
             "session_id": {
               "default": "",
-              "description": "Session id for end.",
+              "description": "End ID or exact ended resume source; same owner/scope, no fallback.",
               "maxLength": 200,
               "title": "Session Id",
               "type": "string"

docs/MCP_TOOLS.md

@@ -25,6 +25,15 @@ class, and execution revalidates availability, scope, authorization, and argumen
 The Smart gateway exposes these nine tools directly; advanced capabilities remain available through
 discovery and the validated executors.
 
+For a deliberate cross-agent handoff, pass the source ID as `session_id` when starting
+`engraphis_session`, or as `resume_from_session_id` to `engraphis_start_session`. The source must
+be an ended session owned by the same authenticated user in the exact resolved workspace and
+repository. Missing, active, deleted, or unauthorized sources fail closed; the tool never
+substitutes a recent handoff. The returned `bootstrap` is
+bounded to 512 regex-counted content tokens, at most six open threads, and per-field character
+limits. `handoff_source` labels the source start/end times in UTC. Sessions have no automatic
+age expiry; those timestamps show age but do not guarantee freshness.
+
 ### Smart routine schemas are reduced by design
 
 The two routine Smart tools deliberately accept smaller allow-lists than their Classic
@@ -183,6 +192,10 @@ an omitted mode means it was not recorded, and is not inferred from current defa
 | Operations | `engraphis_check_update` | Refreshes the release cache and reports whether a newer version is available. Update checks are OFF unless `ENGRAPHIS_UPDATE_CHECK` is set to an affirmative value; `=0` keeps them off. |
 | Decision | `engraphis_decide` | Advisory typed decisions with local fallback. Remote Jev requires an explicit backend and per-call `allow_remote=true`; missing, malformed, and uncertain answers stay visible. Smart discovery routes it through `engraphis_execute_action` because a remote call may consume allowance. |
 
+Classic `engraphis_start_session` accepts optional `resume_from_session_id` for the same explicit,
+same-owner, exact-workspace/repository handoff. Its `handoff_usage` reports the deterministic output
+limit; it is context accounting and does not measure provider billing or task-time savings.
+
 Decision inputs must be nonblank for the selected kind: `guard_command` needs `state`;
 `classify_contradiction` needs `state` and `existing_content`; `verify_support` needs `state`
 and `query`; `verify_completion` needs `state` and `goal`, with optional `recent_actions`.