2.1.0 — quick-wins train: brief load, offline-durable capture, session-end verb

Three independently useful changes (IMPROVEMENT-PLANS #1/#3/#7), no schema changes:

- load --brief: token-budgeted cold-start digest (facts full, last-10 observations,
  scope+freshness, last session's key sections, agent-log lines, inbox count;
  ECHO_LOAD_BUDGET ~8000 chars, lowest-priority-first trimming). SessionStart hook
  injects the brief form; /echo-load keeps the full dump. All load reads (+ the
  sessions listing, which also feeds the heartbeat fallback) fetched in parallel.
- Offline capture durability: a fully-offline capture queues the WHOLE op as one
  semantic record; flush replays it through capture so routing/gate/aliasing re-run
  against the current index; a gate stop on replay is kept + flagged
  (needs_attention, surfaced by flush and load), never landed blind; update-path
  and ensure_daily_log writes ride safe_request; same-args captures dedupe.
- session-end: one call, one lock — session log -> agent-log line -> reflect
  proposals (gate-aware) -> optional scope set -> heartbeat LAST as the commit
  marker; dry-run default; ECHO_NOW pins HHMM; validation runs before any write.
  Stop hook nudge names the command and recognizes it as reflected.
- Fix: main() now catches the __main__ twin-module EchoError (via RuntimeError +
  .code) so helper-module errors exit with their intended codes, not tracebacks.

+19 e2e checks; all suites green. Plans renumbered: MCP container is 2.2.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

README.md

@@ -1,4 +1,4 @@
-# echo-memory — v2.0.0
+# echo-memory — v2.1.0
 
 Persistent memory for Claude / CoWork sessions via the **ECHO** Obsidian vault, driven over the [Obsidian Local REST API](https://github.com/coddingtonbear/obsidian-local-rest-api). The skill makes direct REST calls through a bundled validated client (`scripts/echo.py`). The whole toolchain is **pure Python** (stdlib only), so it runs identically on Windows, macOS, and Linux — no bash, no platform-specific `date`.
 
@@ -421,6 +421,7 @@ From the credential-free harness (`eval/run_eval.py` against the deterministic m
 
 | Version | Highlights |
 |---------|-----------|
+| **2.1.0** | **Quick-wins train: cheaper sessions, durable capture, one-call session end.** (1) **`load --brief`** — a token-budgeted cold-start digest (Fact/Pattern in full, last ~10 Observations, scope+freshness, last session's key sections, Agent-Log lines, inbox *count*; `ECHO_LOAD_BUDGET` default ~8000 chars) now injected by the SessionStart hook, replacing the full six-file dump that grew unboundedly; all `load` reads are fetched in parallel. (2) **Offline capture durability** — `capture` on an unreachable vault queues the *whole operation* as one semantic record; `flush` replays it **through capture** so routing/gate/aliasing re-run against the current index; a gate stop on replay is kept + flagged, never landed blind; `ensure_daily_log`/update-path writes ride the queue too. (3) **`session-end`** — one call (one lock) writes session log → Agent-Log line → reflect proposals → optional scope switch → **heartbeat last as the commit marker**; dry-run by default; `ECHO_NOW` pins the HHMM. Also fixes the `__main__` twin-module trap so helper-module `EchoError`s exit with their intended codes. +19 end-to-end checks. |
 | **2.0.0** | **Packaging & structure major — memory behavior unchanged.** Skills-only: the legacy `commands/` directory is deleted (breaking for pre-skills clients; the 1.6.0-verified skills are the sole entry points). Artifact policy: the repo tracks only the `echo-memory.plugin` pointer; the 15 historical versioned zips leave the tree and versioned builds ship as **Gitea releases** (one per `v<version>` tag) from `v2.0.0` on; `.gitignore` blocks `*.plugin` except the pointer. README gains the "packaging for other agent runtimes" port note (build-target rule, never a second source tree). |
 | **1.6.0** | **Skills-format migration + lint blind-spot fix.** The eight slash commands gain `skills/<name>/SKILL.md` twins (same names, byte-identical bodies; a same-name skill takes precedence, so `commands/` coexists until its 2.0 deletion): per-skill `allowed-tools` end the permission prompts, `disable-model-invocation: true` makes `/echo-sweep` + `/echo-triage` operator-only. `vault_lint` checks retired patterns **before** routes, so a seed README inside a retired tree (`archive/…`) is flagged instead of being absolved by the leaf-README route (+ regression test). `plugin.json` gains `homepage`/`repository`. |
 | **1.5.1** | **Duplicate-gate precision.** Live use caught the gate blocking a decision note because it shared the single vault-common token "echo" with an archived project (min-normalized scoring gives any single-token entity a 1.0 match). New `gate_candidates()` blocks only on **same-kind** candidates (cross-kind name collisions — a decision titled after its project — warn instead), and a **lone shared token blocks only when unique to that entity** (vault df == 1); multi-token overlaps still block. Advisory warnings (`fuzzy_candidates`) unchanged; exit-76/`--merge-into`/`--force` unchanged. +5 tests; verified against the live vault both ways (false positive creates cleanly, true lookalike still gates). |