agentsmesh 0.31.0 → 0.32.0

package/CHANGELOG.md

@@ -1,5 +1,16 @@
 # Changelog
 
+## 0.32.0
+
+### Minor Changes
+
+- 9efbab1: Detect stale files in managed generated-output locations during `agentsmesh check`, and expose `canonicalDrift`, `outputDrift`, and `outputsStale` across CLI, programmatic, and MCP results.
+- a722921: Lessons recall now escalates on repeat failures. When `agentsmesh lessons hook` runs as a `PreToolUse` first-touch guard and the exact action about to run has already failed twice or more (per the opt-in outcome log) with a captured lesson covering it, the covering rule is re-injected above the normal recall bullets as a `RECURRENT FAILURE` escalation — and it cuts through per-session dedup, since a rule the agent saw but did not apply must be shown again. Advisory context only, once per action per session; the covering rule is shown exactly once (the escalation is its delivery), and the default telemetry-off path is unaffected.
+
+  Effectiveness telemetry is now attributed per harness session. The hook stamps the harness `session_id` (and a per-batch relevance `rank`) onto the `delivered` / `failure` rows of `.agentsmesh/lessons/outcome-log.jsonl` and onto recall telemetry, so `agentsmesh lessons stats` groups by real sessions and a failure only impeaches deliveries from its own session — no `AGENTSMESH_SESSION_ID` export needed on hook-driven recalls.
+
+  Added an opt-in `repairTriggers` flag (default `false`) in `.agentsmesh/lessons/config.json` that repairs degraded triggers at capture time: a broad or wide `file_glob` is narrowed toward the evidence file's directory class, a stopworded or over-long keyword gets a matchable variant added beside it, and a keyword that tokenizes to nothing is dropped — each surfaced as a `NARROWED_GLOB` / `KEYWORD_VARIANT_ADDED` / `DROPPED_KEYWORD` warning on the capture result. It never blocks a capture, never rewrites a glob without covering evidence, and never leaves a lesson with zero triggers.
+
 ## 0.31.0
 
 ### Minor Changes

package/README.md

@@ -163,7 +163,7 @@ agentsmesh lessons query --file src/cli/output.ts
 agentsmesh init --lessons && agentsmesh generate   # wire the recall/capture loop once
 ```
 
-`init --lessons` drops a small always-on rule into `.agentsmesh/rules/_root.md` (so every target gets the habit), seeds the full operating manual as a `lessons` skill where supported, and wires a recall hook on hook-capable tools — a `PreToolUse` first-touch guard plus a `PostToolUse` fallback for file/command scope, a `UserPromptSubmit` pass so conceptual (`keyword`) and always-on lessons recall against the **task text**, a `PostToolUseFailure` advisory nudge to capture at the moment something breaks, and a `SessionStart` reset so recall re-surfaces lessons a long chat has summarized away. Agents without shell access use the matching MCP tools (`lessons_query` / `lessons_add`). Because the graph is a normal git-tracked file, a lesson one agent learns today helps every teammate's agent tomorrow, and every change is reviewable like any other diff.
+`init --lessons` drops a small always-on rule into `.agentsmesh/rules/_root.md` (so every target gets the habit), seeds the full operating manual as a `lessons` skill where supported, and wires a recall hook on hook-capable tools — a `PreToolUse` first-touch guard (escalating with the covering rule when the exact action has already failed repeatedly) plus a `PostToolUse` fallback for file/command scope, a `UserPromptSubmit` pass so conceptual (`keyword`) and always-on lessons recall against the **task text**, a `PostToolUseFailure` advisory nudge to capture at the moment something breaks, and a `SessionStart` reset so recall re-surfaces lessons a long chat has summarized away. Agents without shell access use the matching MCP tools (`lessons_query` / `lessons_add`). Because the graph is a normal git-tracked file, a lesson one agent learns today helps every teammate's agent tomorrow, and every change is reviewable like any other diff.
 
 Full walkthrough: [Teach your AI agents with lessons](https://samplexbro.github.io/agentsmesh/guides/lessons/) · [`agentsmesh lessons` reference](https://samplexbro.github.io/agentsmesh/cli/lessons/).